Optics Microstructured Optics SMS Design Methods Plastic Optics

www.led-professional.com ISSN 1993-890X

Review LpR The technology of tomorrow for general lighting applications. Sept/Oct 2008 | Issue 09

LED Driver Circuits Display Report

There are over 20 billion light fixtures using incandescent, halogen, Design of LED or fluorescent lamps worldwide. Many of these fixtures are used for directional light applications but are based on lamps that put Optics out light in all directions. The United States Department of Energy (DOE) states that recessed downlights are the most common installed luminaire type in new residential construction. In addition, the DOE reports that downlights using non-reflector lamps are typically only 50% efficient, meaning half the light produced by the lamp is wasted inside the fixture. In contrast, lighting-class LEDs offer efficient, directional light that lasts at least 50,000 hours. Indoor luminaires designed to take advantage of all the benefits of lighting-class LEDs can exceed the efficacy of any incandescent and halogen luminaire. Furthermore, these LEDs match the performance of even the best CFL (compact fluorescent) recessed downlights, while providing a lifetime five to fifty times longer before requiring maintenance. Lastly, this class of LEDs reduces the environmental impact of light (i.e. no mercury, less power-plant pollution, and less landfill waste). Classical LED optics is composed of primary a optics for collimation and a secondary optics, which produces the required irradiance distribution. Efficient elements for primary optics are concentrators, either using total internal reflection or combined refractive/reflective versions. Secondary optics for homogeneous illumination of circular, square or oblong areas, or line foci is based on e.g., the honeycomb condenser principle, microlens arrays, etc. The design goals are high system transmission, minimum loss of etendue, reduction of straylight and a very short system length compared to conventional illumination schemes. Étendue, as a dominant optical design criterion, is the product or multiplication of the area of the emitter surface and the projected solid angle that the rays from the surface diverge into, and the units are mm2-Steradians. This is a three dimensional version of the Lagrange invariant from imaging or conventional lens design. The losses associated with secondary optics vary depending on the particular element used. Typical optical efficiency through each secondary optical element is between 85% and 90%. Traditional optical design is based on ray tracing or aberration theory. Ray tracing is essentially a sampling technique in which data for a few rays are extrapolated to indicate the performance of an entire system. Aberration theory provides a different type of sampling, in which low-order performance coefficients are balanced with high-order performance coefficients to establish overall performance. The September/October 2008 LED professional Review (LpR) issue highlights LED optics and points out how these technologies and methods can be applied in modern LED lighting systems. We would be delighted to receive your feedback about LpR or tell us how we can improve our services. You are also welcome to contribute your own editorials.

Yours Sincerely,

Siegfried Luger Publisher

Imprint LED professional Review (LpR) ISSN 1993-890X

Publisher Luger Research Institute for Innovation & Technology dep LED professional Rhombergs Fabrik A 6850 Dornbirn, Austria / Europe phone +43 5572 394 489 fax +43 5572 394 698 [email protected]

www.led-professional.com LED professional Review (LpR) - Subsription LpR Digital Magazine – free version Publishing Editors • pdf download Siegfried Luger • view only, low resolution Arno Grabher-Meyer Chrystyna Lucyk • private use LpR Digital Magazine – full version Account Manager annual subscription Tiina Himmer • pdf download • pictures and graphs in high Front-page picture resolutions, printable Mixing Rod/ Optical Research Associates • 6 issues (www.opticalres.com); provided by OEC • access to all previous issues • commercial use Copyrights – Luger Research • Euro 78.80 The editors make every reasonable effort to verify the LpR Printed Magazine – annual information published, but Luger Research assumes no subscription responsibility for the validity of any manufacturers, non profit organisations or individuals claims or statements. • 6 issues Luger Research does not assume and hereby disclaims any • Euro 158.40 liability to any person for any loss or damage caused by errors or omissions in the material contained herein, Send E-mail to: regardless of whether such errors result from negligence, [email protected] accident or any other cause whatsoever. You may not copy, Next LpR Issue - Nov/Dec2008 reproduce, republish, download, post, broadcast, transmit, make available to the public, or otherwise use of LED • LED System Simulation and Testing professional Review (LpR) content without the prior written permission from Luger Research – Institute for Innovation & Technology, Austria.

Content Advertising Index EVERLIGHT p C2 SAMSUNG p 5 Editorial p1 EDISON p 13 Imprint p2 AUSTRIAMICROSYSTEMS p 16 Project News p4 SPHEREOPTICS p 17 Product News p6 JENCOLOR p 17 IP News p12 ALPHA-ONE ELECTRONICS p 17 LED TAIWAN 2009 p 21 LED CHINA 2009 p 27

Application INSTRUMENT SYSTEMS p 40 Progress in LED and OLED Display Technologies ROAL LIVING ENERGY p 51 by M. Nisa Khan, Ph.D., LED Lighting Technologies p14 POWER VECTOR p 51 SOLID STATE LIGHTING ASIA p 52

Optics INTERNATIONAL RECTIFIER p C3 Microstructured Optics for LED Applications TRIDONIC.ATCO p C4 by Arthur Davis, Reflexite Optical Solutions Business p18

Review of SMS Design Methods and Real World Applications by Oliver Dross et al., Light Prescriptions Innovators LLC p33

Plastic Optics Enable LED Lighting Revolution by Tomi Kuntze, LEDIL Oy p41

Driver New Drive Circuit Supports Wide DC Input and Output Range by Bernie Weir, ON Semiconductor p44

Simple and Accurate Constant Current Sensing for Non-Isolated LED Drivers by John S. Lo Giudice, STMicroelectronics p47

Reliability Test of LED Driven by PWM Technique by B. J. Huang et al., National Taiwan University p49

Project News

Semperlux and Vossloh- Schwabe Illuminate High Water Pumping Station Cologne has been graced with a new landmark: the high-water pumping station in the Bayenthal suburb of the city. The pumping station is already a sight to see during the day thanks to its architectural aesthetic, which Different appearance of the pumping station from a distance view . includes an organically landscaped area that inconspicuously tucks the To suit this specific project, a system was developed that involved structure into the banks of the Rhine. The control building, which was encapsulating standard VS Optoelectronic Flex modules fitted with SMD designed by the architectural company Kasper Kramer (BDA) to feature a LEDs inside the aluminium sections of the metal grid façade. The white, red, contemporary metal grid façade, was kept prominent as the visible face green and blue Flex line modules, each backed with self-adhesive tape, were of the facility. The breathtaking illuminations concept developed by the first precisely trimmed to fit and then fixed within the inside face of the Semperlux lighting technology specialists in Berlin ensures the pumping respective segment. The LED modules were also specifically adapted to station lights up in one of three colours depending on the water level of ensure even light-point spacing and uniform light density over the entire the Rhine. A normal watermark is indicated by white light, rising water building. With the fitting order of the modules having been precisely levels by amber and on reaching the high water mark, the light turns red. calculated by the light engineers at Semperlux, actual installation was The station then begins pumping off the discharge in the reinforced problem-free thanks to the highly flexible nature of the LED modules. conduits. As a result, the citizens of Cologne are not only aware of when Finally, the standard Flex modules were permanently encapsulated within the pumping station is at work, but can also enjoy the eye-catching light the metal sections using a transparent potting material to ensure protection spectacle as a thing of simple beauty. against environmental factors. The indirect LED lighting system works by light hitting the walls of the building and then being reflected to appear as Vossloh-Schwabe Optoelectronic provided the LED technology that made an evenly illuminated surface. In response to the actual water level of the the artistic aspirations of this unique illumination system possible. Project Rhine, a DALI unit triggers the respective light scene and a EUTRAC® DALI decision makers opted for the LED solution for several reasons. Next to low LightComposer®-REG module enables overall system control. As a result, energy consumption, high light output and flexible light control (thanks to both the citizens of Cologne and visitors to the city can enjoy these special RGB technology), LEDs guarantee an extremely long service life, practically illuminations night after night. no maintenance and, with that, very low life-cycle costs.

The pleasing daytime look of the high-water pumping station is only topped by the ‘wow factor’ that nightfall brings.

www.led-professional.com LED professional Review | Sept/Oct 2008 | page 4

Reliability is guaranteed as IST offers a 5-year warranty, making the new Product News DL-0x0 range extremely attractive to high-end developers, high profile public buildings, and exclusive hotels. For applications where lighting is used for approximately 12 hours per day, the return on investment is estimated at 2-2.75 years. For applications with 24 hours operation, the IST Launches High Efficient return on investment is estimated at 1-2 years. The variables are dependant LED Downlights upon the volume of downlights per application. Integrated System Technologies Ltd, a UK designer / manufacturer of LED systems and the industry leading iDrive range of LED drivers, has launched the first high-powered LED downlight range that is capable of Sharp to Introduce 11 LED replacing CFL fixtures. The patented DL-0x0 LED downlights are available in cool 5500K CCT and neutral 4100K CCT colour temperature options. Luminaries This range of fixtures offers outstanding performance and a direct Sharp Corporation has developed a series of new LED Lightings, alternative to a twin 26-watt compact fluorescent downlight. including the “oblong” type that features a brightness equivalent to the fluorescent lamp fixtures*1 that are currently the main lighting in factories, offices and commercial spaces, and the “downlight” type with a brightness equivalent to a standard 150-watt incandescent light lamp. Sharp will introduce a total of 11 models into the Japanese market, including four oblong, one square, and six downlight models.

Patented DL-0x0 LED downlights. Some of Sharp’s new LED luminaries.

The light engine lamp lumens for the DL-090 was independently Today, looking at a breakdown of the amount of electric energy measured at 1060 lumens and the DL-080 at 963 lumens. These figures consumed by lighting in Japan, business and commercial applications were calculated after the light engine had been running for 1 hour and typified by factories and offices consume roughly double the amount measured in an integrating sphere. The LOR (Light Output Ratio) for the used for residential home use. Thus, there is increasing demand for system is 0.89, which exceeds any traditional lighting product on the replacement lighting options with a view toward switching to the next market. The power consumption is around 20 circuit watts which means generation of lights that offer high environmental performance and the DL-080 downlight exceeds 40-lumens-per-watt system efficiency which do not use hazardous substances such as mercury. and the DL-090 exceeds 45-lumens-per-watt system efficiency. For the highest output options, the average Lux levels for the DL-080 are 2170 LED lights, which have low energy consumption compared to at 1m, 545 at 2m and 242 at 3m. The highest option DL-090 provides incandescent lamps and fluorescent tubes, and which feature 2270 x at 1m, 567 x at 2m and 252 x at 3m. There are three cover outstanding environmental performance including a long product life options which include the standard diffuser, drop, and an option without and being mercury-free, are expected to serve as the next generation of diffuser. In addition, gold and chrome bezel finishes can be ordered lighting sources. Sharp has been involved in the development and instead of the standard white finish. commercialization of LED devices for more than 30 years. Sharp has been working to take full advantage of its know-how nurtured over The DL-0x0 range has been developed to offer a true alterative to a twin that time to enhance the brightness of these devices and assemble 26-watt CFL downlight. As a starting point, IST purchased and independently them into modules in order to develop new lighting fixtures, undertake tested a twin 26 watt CFL downlighter at the Photonics Cluster (UK). It was mass production, and bring new LED Lightings to the marketplace. found that the CFL consumed around 68 watts of input power. The lamp lumens were rated at 3600. When tested in an integrating sphere, The “oblong” lightings being introduced at this time feature an elongated it only emitted 1750 lumens. This meant the true LOR was 0.49. In form factor and deliver brightness equivalent to conventional fluorescent addition, IST found that the LUX levels were 2000 at 1m, 500 and 2m light fixtures equipped with twin 40-W straight-tube fluorescent lamps and 200 at 3m. These figures were used to produce an LED downlight thanks to a luminaire efficacy of 74 lumens/watt, the highest in the equivalent. industry for LED lighting. They also offer a high level of energy efficiency,

consuming about 25% less power than conventional fluorescent Colors are evenly mixed when leaving the fixture, making for an fixtures. Both the oblong model and its companion “square” type model, attractive looking luminaire. The Alien LED Downlight is 0-100% which is ideal for conference rooms and reception areas, provide even, dimmable for balance of brightness or the ability to respond intelligently uniform illumination thanks to diffused surface-emitting light to the availability of natural light. technology. Further, the “downlight” line-up includes a model with light The Alien LED Downlight is available in two versions to cover a wider output equivalent to a conventional 150-watt incandescent bulb, the range of applications; a 9 W standard version (5 LEDs) and an 18 W highest brightness in the industry for LED Lightings. These downlight high-power version (9 LEDs). Very bright when compared to other models are suggested for use in commercial spaces such as shopping products in its class, well-known Rebel LEDs are used to ensure high centers and department stores, as well as hotel lobbies and entrance performance and output as well as a high quality. Because lamp foyers to business offices. replacement isn’t every 3000 or 5000 hours, no service is required other LED Lightings manufactured by Sharp will be adopted as the main than regular cleaning. lighting in all plants to be located within the “Manufacturing Complex For even greater flexibility in design, two optical systems are available, for the 21st Century” now under construction in Sakai City, Osaka medium or wide, covering distances from approximately 0.5 to 3 meters Prefecture, and scheduled to begin operations by March of 2010. This (2 to 9ft). will represent the world’s largest collection of LED Lightings installed in buildings on this scale. Using this achievement as a springboard, Sharp The required Alien LED Driver, sold separately, is a plug-and-play unit will be working in the future to further expand its LED lighting business that offers great flexibility for solutions using both 9W and 18W fixtures. into factories, offices and commercial spaces around the globe. One Driver can control up to a maximum of 20 standard luminaires or 10 high-power luminaires, or a combination of the two. The Alien LED Downlight will fit into any standard MR16 hole, allowing it to blend inconspicuously into existing architecture and making it New Alien LED Downlight™ ideal as a replacement product or for new installations. Alien LED from Martin Architectural Downlights are easily interconnected by linking transformers together with standard IEC connectors. The Alien LED Downlight™ is a plug-and-play recessed LED downlight that features RGB+W color mixing and is available in a standard or high The Alien LED Downlight has two DMX control modes, HSI & RGBW for power version. Designed for accent lighting in colored or variable white comprehensive control when a designer needs it. The intuitive HSI light, the Alien LED Downlight’s IP67 rating means it’s equally suited to control system allows designers to control the unit in terms of color both indoor and outdoor environments. rather than only DMX levels. A stand-alone feature offers 31 pre- programmed static or dynamic sequences via the Alien Driver Unit. The Alien LED Downlight is made of a durable steel and aluminum construction and the IP67 rating means it’s fully weather-protected. It comes with stainless steel trim rings as standard with chrome, white or brass available as accessories. The luminaire is convection cooled so no fans or other moving parts are

The new Alien LED Downlight™ plug-and-play recessed LED downlight. used. Rigorously tested and CE, ETL & CETL approved, the Alien LED Downlight is built for simplicity. The Alien LED Downlight can be mounted on walls or ceilings and is ideal for expressive background lighting in restaurants, cafés, bars, small clubs and lounges. It is useful for wall grazing, as decorative accent lighting and niche lighting, or for bottle or product displays. The Tri-O-Light Casting Resins: Alien LED Downlight is also perfect for entrance lighting or orientation lighting for marking columns, aisles, walkways, steps, and architectural Freedom to Sculpture features. The light line in the street in front of the entrance to the Gelredome stadium (Arnhem, The Netherlands) is one of the projects which have The Alien LED Downlight’s RGB+W diodes excel at creating a broader been realized with the LED Light Strip RGB of Tri-O-Light and the range of hues including deep, saturated colors. The addition of the polyurethane casting resin technique of Multicol casting adhesives white LED gives several advantages: a true white, the ability to adjust trading: a new expertise within Tri-O-Light. In addition, the casting color temperature, and the possibility to create soft pastel shades. resins were used as FLEXLED, indirect LED lighting underneath benches in a shopping centre in Amsterdam, and the bus stop “Arena Boulevard”, as well as curbstone lights on several roundabouts in the Netherlands.

Conformed to the wishes of the client, a flexible, diffuse RGB LED-light consumption, the LED High Bay provides a compelling business case for line has been developed, which is water resistant, flexible, durable and one-for-one replacement of metal halide or high bay fluorescent of high impact. The colour of the 60m-long light line can change and is fixtures. The economics of Constellation can be further enhanced with controlled from the reception area of the stadium. For every event, the the use of the optional motion sensor or off-the-shelf daylight correct colour adjustment is shown using the DMX control panel. The sensors. light line was supplied in six prefab parts and glued into the ready made compartment. The flexibility of the material allows the light line to be inserted in a radius. Multicol is the right partner for the production and distribution of a large selection of highly developed types of adhesive for the industry. It also handles pouring products in a casting resin for numerous applications.

Project examples for Tri-O-Light casting resins. The Constellation series LED High Bay luminary.

In a state-of-the-art equipped lab, Tri-O-Light disposes of the newest “Solving our customers’ energy, maintenance, and environmental casting resins, such as transparent polyurethane casting resins. They problems drives Albeo’s creative process,” said Jeff Bisberg, CEO. “Our pour fixtures to the customer’s requirements. In this chosen fixture, LED deep electro-optical design capabilities have allowed us to lead in the lighting is built. Thanks to the advanced pouring process, the customer industrial LED marketplace with the first dedicated industrial high bay or the designer has the possibility to develop products following their fixture. Our new trial program now enables customers to test this own design. All casting resins are solvent free, transparent and extremely leading-edge solution and accurately evaluate the business case for a suitable for outside applications. switch to LED lighting.” The properties of the polyurethane castin resin can be adjusted by Tri- In addition to best-in-class performance and lifetime, Constellation O-Light’s specialists, ranging from very hard and rigid to very flexible. High Bay customers will be able to enjoy the many associated benefits The colour can also be adjusted, from transparent to diffused. And the of LEDs, including: resin itself can also be coloured. • LEDs run cooler than traditional fixture types, reducing air conditioner loads • LEDs are fully dimmable and compatible with various controls ALBEO LED High Bay Delivers • LEDs contain no hazardous mercury, so they do not require recycling • LEDs are nearly unbreakable Maintenance-Free Lighting • LEDs are free of annoying flicker and buzz • LEDs are safer, operating on low voltage DC, with no glass or vacuum Albeo Technologies released its Constellation series LED High Bay luminaire for industrial, commercial, warehouse and retail spaces. Albeo Because LED lighting is new to general illumination, particularly high is the first LED light fixture manufacturer to make an LED fixture that bay applications, Albeo has created a trial program for users to test the combines the reliability of LED lighting with the fixture performance fixtures in their own building. The Albeo trial program allows customers requirements for high bay applications. to test a sample of fixtures for 60 days with no risk. Maintenance of high bay lighting is expensive: the lamps don’t last long; they require a lift to service; and they are frequently placed over operating machinery, public areas, or sensitive material. The high cost of changing lamps and ballasts puts a premium on fixture longevity, and now the Constellation LED High Bay provides high bay performance with over 50,000 hours of operating life. When combined with Constellation’s 20,000 lumen output and 360 watt maximum power

New Polycarbonate Optics Cree: New XLamp® XP-E and from Carclo XP-C LEDs Carclo Technical Plastics announces the immediate availability of a wide Cree, Inc. announces a new standard for lighting-class LEDs with the range of high performance polycarbonate optics to manage the light introduction of the XLamp® XP-E and XP-C LEDs. These breakthrough output from new Cree XLamp® XP-E and XP-C LEDs. LEDs have the smallest footprint in the industry for lighting-class LEDs— providing the same high-quality lighting performance and proven These new optics are available immediately in 10mm square “small reliability as Cree XR-E and XR-C LEDs in an 80% smaller package. footprint” optics, and soon in standard 20mm and 26.5mm diameter sizes, in both a clear and a frosted finish. The 10mm optics feature a The new XLamp LEDs, measuring just 3.45mm square by 2mm high, can clear medium beam with a FWHM (Full Width Half Maximum) beam enable new applications, including backlighting, signage, outdoor, angle of 16.4 degrees, an elliptical or “line” optic with a 43x16 degree indoor and portable lighting, thanks to their small size and low profile beam, as well as Carclo-exclusive frosted optics in medium (18 degrees as well as a wide viewing angle and symmetrical package. Available bins FWHM) and wide (26 degrees FWHM) beams. Carclo 10mm “small for XLamp XP-E LEDs include minimums of 100 lumens at 350mA in footprint” optics do not require a separate holder, and can easily be cool white (5000K - 10000K) and 80.6 lumens at 350mA in warm white grouped together in square, rectangular or linear multi-LED arrays. (2600K - 3700K).

Carclo 10mm optic and model of multi-LED array.

Carclo also wishes to announce the immediate availability of a 20mm side-emitting optic for the Cree XLamp® XP-E and XP-C LEDs, with an industry-leading 5-degree beam angle around a full 360 degrees, and efficiencies approaching 90%. Finally, 20mm and 26.5mm optics for the new Cree XLamp® XP-E and XP-C LEDs will soon be available in narrow, medium and wide beam angles, in both clear and Carclo- exclusive frosted optics, as well as elliptical and elliptical orthogonal line optics. Ian Bryant, Business Development Manager for Carclo’s UK optics business, said: “Cree XLamp® XP-E and XP-C LEDs have been developed by Cree for lighting-class performance in a small-footprint, low-profile Small footprint high power LEDs: XLamp® XP-E and XP-C LEDs. LED package, and to give LED lighting designers enhanced flexibility and performance to further accelerate the LED lighting revolution. We are “We recognized an unmet need for lighting-class performance in a delighted to be able to offer a line of optics to coincide with the launch small-footprint, low-profile LED package. These products, based on an of these new LEDs.” Jim O’Connor, Business Development Manager for innovative new technology platform, address this need,” said Paul Carclo USA added: “Carclo’s timely announcement of a wide range of Thieken, Cree marketing director for LED components. “This new optics for the newly introduced Cree XLamp® XP-E and XP-C LEDs is a platform, in concert with the existing XLamp products and the recently further indication of Carclo’s firm desire to support users of new LEDs demonstrated XLamp MC-E LED, give LED lighting designers enhanced developed by all the major manufacturers. We believe that our unique flexibility and performance to further accelerate the LED lighting 10mm small-footprint optics and our Carclo-exclusive frosted optics revolution.” will provide the kind of performance that will allow these new LEDs to be rapidly incorporated into new designs for LED light engines and solid state lighting fixtures.”

The MAX16822A/B and MAX16832A/C integrate thermal-shutdown Maxim: High-Efficiency circuitry to protect the driver under excessive thermal conditions. Their 200mV, high-side current-sensing minimizes the voltage drop across HB LED Drivers for Greener the external sense resistor, thereby reducing power dissipation. Additionally, they provide analog thermal-foldback protection that Lighting Solutions reduces the LED current in conjunction with an external negative Maxim Integrated Products introduces the MAX16822A/B and temperature coefficient (NTC) thermistor when the temperature of the MAX16832A/C family of 2MHz, high-brightness (HB) LED drivers with LED string exceeds a specified limit. an integrated MOSFET and high-side current-sense circuitry. These This robust family of HB LED drivers operates from a 6.5V to 65V input highly integrated devices reduce the size, complexity, and cost of solid- voltage and is fully specified over the -40 degrees Celsius to +125 state lighting (SSL) designs, simplifying the implementation of green degrees Celsius automotive temperature range. The MAX16822A/B are lighting technology. Moreover, they provide the high conversion available in a lead-free, 8-pin SO package, while the MAX16832A/C are efficiency and advanced thermal-management capabilities critical to offered in an 8-pin SO package with exposed paddle. Prices start at LED system designs. The MAX16822A/B and MAX16832A/C are ideal $0.99 for the MAX16822A/B, and $1.31 for the MAX16832A/C (1000- for MR16, MR111, and other architectural lighting applications, as well up, FOB USA). as front and rear automotive lighting (RCL, DRL, and fog/low-beam lights). These high-voltage, switch-mode HB LED drivers are capable of delivering over 18W (MAX16822A/B) or 36W (MAX16832A/C) of output 100V, Full Featured LED power with up to 95% efficiency, while driving as many as 15 white LEDs in series. The MAX16822A/B can deliver up to 350mA of continuous Controller for High Current DC current, while the MAX16832A/C can deliver up to 700mA of continuous DC current to the LED string. These devices offer an analog LED Applications dimming feature that allows the LED current to be controlled by an Linear Technology announces the LT3756, a 100V, high-side current external DC voltage and PWM dimming input for a wide range of sense DC/DC converter designed to drive high current LEDs. Its 6V to brightness levels. 100V input voltage range makes it ideal for a wide variety of applications, including automotive, industrial and architectural lighting. The LT3756 These HB LED drivers require only a few external passive components uses an external N-channel MOSFET and can drive up to 20 1A white for the design of a low-cost LED driver circuit. By integrating a 65V LEDs from a nominal 12V input, delivering in excess of 70 watts. It switching MOSFET with RDSON values of 0.85ohms (MAX16822A/B) incorporates a high-side current sense, enabling it to be used in boost, and 0.45ohms (MAX16832A/C), they eliminate the need for an external buck, buck-boost or SEPIC and flyback topologies. The LT3756 can power transistor. Additionally, high-side current sensing and integrated deliver efficiencies of over 94% in boost mode, eliminating any need for current-setting circuitry further reduce the number of external external heat sinking. A frequency adjust pin permits the user to components required, while delivering ±3% LED current accuracy. program the frequency between 100kHz and 1MHz, optimizing efficiency while minimizing external component size and cost. Combined with a 3mm x 3mm QFN or a thermally enhanced MSOP-16 package, the LT3756 offers a very compact 70-watt LED driver solution.

Typical Operating Characteristics (VIN = VDIM = 48V, CVCC = 1μF, RSENSE = 0.62Ω, L = 220μH;@ TA = +25°C).

A hysteretic, step-down (buck) mode enables these devices to drive a long string of HB LEDs from input voltages as high as 65V DC. This control method renders control-loop compensation unnecessary, thus simplifying the design and minimizing component count. The MAX16822A/B and MAX16832A/C are designed to hold the LED current ripple to within 10% (A versions) and 30% (B and C versions) without the use of an output capacitor. Moreover, the devices support switching frequencies up to 2MHz, allowing the use of small-sized components. Design example for 90% Efficient, 20W SEPIC LED Driver.

The LT3756 uses True Color PWM™ dimming, delivering constant LED color with dimming ranges up to 3,000:1. For less demanding requirements, the CTRL pin can be used to offer a 10:1 analog dimming range. Its fixed frequency, current mode architecture ensures stable operation over a wide range of supply and output voltages. A ground referenced voltage FB pin serves as the input for several LED protection features, making it possible for the converter to operate as a constant- voltage source.

Summary of Features: • 3000:1 True Color PWM Dimming • Wide Input Voltage Range: 6V to 100V • Output Voltage Up to 100V • Constant-Current & Constant-Voltage Regulation • 100mV High Side Current Sense CAT4101 block diagram. • Drives LEDs in Boost, Buck Mode, Buck-Boost Mode, SEPIC or Flyback Topology Since it is a linear-based LED driver, the CAT4101 does not require an • Adjustable Frequency: 100kHz to 1MHz inductor, eliminating noise, minimizing component count and simplifying • Programmable Undervoltage Lockout with Hysteresis design. The LED current is set via an external resistor connected to the • Open LED Status Pin (LT3756) RSET pin. The LED pin is compatible with high-voltages up to 25V. This • Frequency Synchronization (LT3756-1) enables the CAT4101 to drive long strings of up to 10 LEDs, adjustable • PWM Disconnect Switch Driver to 1A. The device offers high-speed, high-resolution PWM dimming, an • CTRL Pin Provides Analog Dimming “instant-on” PWM control mode and thermal shutdown protection in • Low Shutdown Current: <1uA the event of an over-temperature fault condition. The CAT4101 is • Programmable Soft-Start available in a 5-lead, TO-263 package to provide exceptional thermal • Thermally Enhanced 16-Lead QFN 3mm × 3mm Package or MSOP- dissipation. 16E Package

Two versions of the LT3756 are available. The standard LT3756 offers an Open LED Status pin, and the LT3756-1 replaces the Open LED Status Ultra-Large 8” Sapphire pin with a frequency synchronization pin. Wafers for LEDs Monocrystal, a leader in the synthetic sapphire and other advanced electronic materials markets, has announced that the company launched Catalyst: 1A Constant- production of its ultra-large 8” C-plane epi-ready sapphire wafers for Current, Low Dropout Driver LED manufacturing in August. Catalyst Semiconductor, Inc. a supplier of analog, mixed-signal and non-volatile memory semiconductors, has expanded its line of constant- current, LDD™ (low dropout driver) products to include a new, simple- to-use, high-power device for architectural, landscape, automotive and general LED lighting products. The CAT4101 is a linear-based, constant- current LED driver, which drives long strings of LEDs and provides a high-current solution for a wide range LED applications. Product Features: • 1A current sink with high accuracy • Up to 25V operation on LED pin • Low dropout voltage drive 500mV/1A • LED current set by external resistor • High resolution PWM dimming via EN/PWM • Thermal shutdown protection • Packaging: 5-lead TO-263 The 8” sapphire wafer has 16 timest the face of a standard 2” wafer.

Monocrystal was able to rapidly ramp up production of the new The SPR’3 system’s scope of application includes the measurement of generation sapphire wafers for LEDs by leveraging its superior light sources, lamps and optical systems within a wavelength band of technology for growing large sapphire crystals – the company has 200 to 1100 nm. The measurement of the absolute spectral emission routinely produced large sapphire crystals that exceed 65 kg since late provides very high accuracy for the measurement of monochromatic 2005. Another key factor that helped Monocrystal advance on 8” wafers light sources, as well, for example with LEDs. The measuring scale can for LEDs was its proprietary wafer fabrication technology developed even be enlarged and radiation prefiltered by an integrated filter wheel earlier in spring for production of the new generation ultra-large 8” using any filter you choose. A Spectrometer and absolute sensors form R-plane sapphire wafers for RFICs. Monocrystal already shipped its 8” a compact unit. A USB interface provides a simple data connection to wafers to LED and RFIC customers. any computer or laptop. “Being first to market with our ultra-large 8” wafers for LEDs is a clear The total calibration information is stored on the sensor head. Thus the indication of Monocrystal growing brand’s exposure and recognition, system can be operated by different computers. The robust system its expanding market and technology leadership,” stated Vladimir works without external optical fibre and, therefore, can also be used in Polyakov, Chairman of the Board of Monocrystal and President of a rough environment. The control and evaluation software was Energomera, the mother company of Monocrystal. “To be successful in redesigned and offers a variety of analysis methods. Emission, reflection a hightech market, you should truly possess the capabilities of a and transmission measurements round off the scope of performance of technology leader, and be able to constantly innovate and offer products the SPR’3 measurement system. of the highest quality that meet the current and future needs of your opsira GmbH delivers the SPR’3 unit completely calibrated and also clients. Monocrystal exports 90 percent of its production, and is a key provides the recalibration service. supplier for LED and RFIC markets,” he concluded. Today, increasing demand for energy-efficient and “green” materials, products, and technologies is becoming a solid global trend. The introduction by Monocrystal of its ultralarge diameter sapphire wafers IP News for solid state LEDs is a socially responsible contribution of the company to address growing global concerns of inefficient energy consumption, carbon emission, and use the of hazardous materials. OSRAM and Philips Conclude LED Cross-License Agreement opsira: Spectroradiometer LED luminaire manufacturers can now use key components from OSRAM without having to pay license fees to Philips because OSRAM SPR’3 has acquired special rights from Philips. Luminaire manufacturers are The new conceptional design of opsira’s spectral measurement systems therefore released from the costs resulting from the LED luminaire is aimed at developing a portable system for absolute spectral emission license program recently published by Philips. measurements from UV to NIR. The special rights relate not only to patents held by Philips but also to patents held by Color Kinetics and TIR Systems which were acquired by Philips last year. The compensatory payment – which both parties have agreed not to divulge – will be partially offset by OSRAM by selling rights to its own patents. This agreement will give the market for LED- based lighting a further boost. “Thanks to this agreement, OSRAM key components provide our customers with unique opportunities. It stimulates new possibilities for the lighting market and, like Philips, we will accelerate the development of this market”, said Martin Goetzeler, CEO of OSRAM. The LED components are based to a large extent on chips fabricated in OSRAM’s highly advanced factories in Regensburg and Penang/Malaysia, and this is an area in which OSRAM itself has a leading patent portfolio.

Portable Opsira Spectroradiometer SPR’3.

www.led-professional.com LED professional Review | Sept/Oct 2008 | page 12

The OLEDs on the other hand are fabricated using a method similar to Application ink-jet printing. This method can precisely deposit, with sufficient isolation, organic RGB molecule dye chemicals or predetermined emissive polymers into miniature wells fabricated in organic layers. The Progress in LED and OLED RGB well combination forms a pixel in passive or active matrices known as PMOLED or AMOLED displays that can provide beautiful colors. The Display Technologies repeatable chemical fabrication process along with monolithic integration of organic layer transistors (TFTs) allow for superior OLED > M. Nisa Khan, Ph.D., President, LED Lighting Technologies color display and video processing. One of the latest OLED breakthroughs No one doubts that numerous challenges are preventing LEDs from has been a 31-inch TV prototype from Samsung SDI shown in CES 2008 becoming ubiquitous in the general illumination market. However, the in Las Vegas; this 1080p panel, just 4.3mm thick, uses half the power of illumination industry leaders see this as a possibility over a 10-year a “typical” 32-inch TV. Their 2005-version was a 21-inch OLED display horizon, if not sooner. Already LEDs have overtaken in traffic lights, featuring the highest WUXGA resolution with 6.22 million pixels. They some automotive head and rear lights, channel letters in signage, and both adopted the AMOLED technology for its low power consumption in some architectural lighting applications. In most of these applications, and high-resolution qualities. The screens have brightness of 400 nits, inorganic III-V compound semiconductor LEDs are favored and OLEDs contrast ratio of 5000:1, color gamut over 75% and fast response times, are considered less-mature counterparts, at least for some time. making them ideal for viewing HD-resolution video images. The best AMOLEDs can display almost four times as many colors as equivalent- While LEDs are not yet the overwhelming choice in various lighting and size LCDs can produce with much wider viewing angles. other illumination markets, they are very popular in billboards and various other large-format indoor and outdoor displays. Their large Samsung SDI (Figure 1) also boasted recently about their 12-inch viewing distances allow for coarse resolution or pitch, which can be AMOLED display-fitted notebook possessing a 1,280 x 768 resolution achieved with individual RGB packaged-LEDs placed in proximity. These with all the usual benefits of this display technology. They claim a LEDs are mostly the more mature inorganic III-V semiconductor LEDs. contrast ratio 20 times that of conventional LCD displays! However, when it comes to smaller screens requiring much higher resolutions as in consumer electronics namely cell phones, cameras, various other handheld devices, prototype laptops and television screens, OLEDs have been the choice over the III-V LEDs. Why? The reason is that a planar array or uniform, high-quality LEDs in very close proximity needed to form a high-resolution viewing screen is not possible with III-V semiconductors over too large an area. The reasons are multifold. First, high-quality RGB LEDs need to come from different III-V compound materials and therefore monolithic integration of 3-color pixels in a single flat plane is not feasible. A hybrid integration of inorganic RGB LEDs with 16micron pixel pitch with adequate color isolation (16 micron pitch is needed to achieve 160 pixels per inch, required in 1920X1200 WUXGA laptop screen; 1micron = 1/1000cm) is surely a daunting task. Even if these RGB pixels could be constructed in a III-V plane, it would be limited by the 2-inch standard wafer size in Figure 1: Samsung Demos OLED Display-Fitted Notebook - Photo: Courtesy www.electronista.com (May 16, this industry. As of now only a few high-end LED manufacturers have 2008). 3-inch or 4-inch wafer processing capability, which is still rather small for display technologies. Furthermore, single-color LEDs processed in Despite the OLED promises, laptops, most cell phones, and other hand- a wafer are subjected to such significant challenges as non-uniformity, held display devices predominantly use LCDs, a technology invented in impurities, dislocations, and lithographic variations, all of which lead to 1963 with a hope to replace bulky CRTs. While OLEDs need further the well-known ‘binning’ issue. These challenges are particularly improvement in lifetimes, LCDs are difficult to scale up to large surfaces. pronounced at the edge of the wafers, making a good portion of a LCD production and commercial expenses are still high and therefore wafer unusable. Even if III-V RGB LEDs, say, could be produced over an remain vulnerable to new innovations. area as large as one square centimeter using some flip-chip wafer Samsung, SONY, Osram, and Kodak are but a few big companies banking bonding technique, tiling them to create a small cell phone screen is on OLEDs for various display applications. ISuppli Corporation believes also unlikely because of an evitable tiling gap of at least 1/100 of an by 2010 factories will be churning out 289 million AMOLED displays inch would still be unacceptable. annually estimating that about 88% of them will end up in cell phones.

Today OLEDs are commercially used in limited quantity cell phones, cameras and car stereo displays because they are only on part of the time. Using them in computers and TVs may take another five years as companies work on developing longer-lasting chemicals. Large-format displays viewed from much larger distances than those typical of consumer electronic devices can use larger pixels and lower resolutions. Inorganic LEDs are best-suited for them because these bright, robust, long-lasting and high-quality LEDs can be packaged as surface-mounts and be integrated on a mother-board with drive electronics, with a pixel pitch as little as 4 mm. Such pitch-size or greater is appropriate for large-screen indoor and outdoor displays found in billboards, stores, malls, theatres, and on-premises of town plazas. High-brightness combined with dynamic color and resolution options make these large LED displays superior to other display technologies.

One must however regulate a number of elements to maximize viewing Figure 2: LED Billboard by QS-Tech delivered to the National Stadium in Beijing, China (Photo: Courtesy of QS- quality when using LED display panels. These are screen resolution, Tech). brightness and contrast, video processing, electrical usage and When pixels are viewed up close, the RGB LEDs appear as separate dots. maintenance. These LEDs mix to form a single color at a distance from the screen known as the color compound distance (CCD). For indoor displays, Screen Resolution more advanced color compound scheme is required to make images Resolution is the key to a quality image on a large-format video display. appear clear and sharp from close range. Best indoor displays use 3-in-1 It is defined by the total number of vertical and horizontal pixels that RGB in one packet with SMD LEDs, where as outdoor displays use form the picture. The picture on the screen is typically reproduced from individual RGB LEDs. Typically, CCD is calculated by multiplying the pixel a video signal that have a native NTSC resolution of 486 (vertical) X 240 pitch in millimeters by 250mm for indoors and by 500mm for outdoors. (horizontal) or 576X720 (PAL). A screen with fewer pixels than the The minimum viewing distance is determined by multiplying the pixel input video signal will have less resolution than the source; for best pitch by 750 to 1,000. The maximum viewing distance is generally 20X results, a minimum resolution of about 648×486 (NTSC) or 768×576 to 30X the screen height. (PAL) is recommended; however, a properly designed display with lower resolution can still give an acceptable appearance for video images, like Brightness & Contrast those seen in a standard VGA screen. Taking this as a benchmark, to Brightness, or more accurately luminance, for LED screens is measured achieve the same 640×480 VGA (15-inch diagonal) resolution with a in nits (cd/m2) using a light meter. The general rule is to require at least 3m × 2.25m screen, the pixels would need to be spaced approximately 1,000nits for an indoor display and 5,000nits for outdoor displays. The 4.5mm apart. This distance between pixels is called pixel pitch, measured designer first measures a full white signal at a normal angle to the usually in millimeters. screen, setting color temperature at 6,500K (D65) for outdoor screens, Typically, indoor-screen pixel pitches are 6mm, 10mm, 16mm, and and 5,000K (D50) for indoor screens. He then repeats this measurement 20mm, whereas 16mm, 20mm, 25mm, and 30mm are used for outdoor several times at various points (equally spaced) of the screen. Setting screens. Outdoor screens tend to be larger than indoor screens because the screen to black, he then re-measures for the ambient reflected light. the viewing distance is often greater. QS-tech (QS-tech.com), a Chinese The brightness is an average of the various points of white, minus the manufacturer, now offers pixel pitches starting from only 4.4 mm for measured ambient when the screen is black. indoors and 10mm for outdoors. They will be one of the major suppliers The viewing angle (VA) is defined at the point when brightness falls to for the 2008 Olympic Games in Beijing. Fig. 2 shows an LED billboard 50% from the maximum value. By walking around the screen, the (6-mm pixel pitch) they delivered to the National Stadium in Beijing for designer can determine how brightness changes, and review the RGB the 2008 Summer Olympic Games. and white to see if the colors remain uniform at all angles. A challenge Larger pixel pitch leads to higher pixelization meaning the viewer starts unique to LED display technology is known as shouldering, a color shift to see the pixel structure. Hence, the screen designer needs to calculate caused by one LED blocking the view of another at extreme angles. The it based on the distance between the viewer and the screen. The choice VA should include color shifts as well as brightness; if a significant color is also dictated by any physical size constraints, sight lines, and of shift occurs at an angle before the brightness falls by 50%, then VA is course, budget. reduced to this angle.

Screen manufacturers should not drive the LEDs at high currents and flicker-free video, generally done by dedicated video processing quote impressively high brightness figures. High drive currents lead to equipment. A standard step is to view the content first using a CRT TV faster degradation in the LEDs, leading to dramatic screen non- monitor, although it is not always a true representation of how the uniformity. Quoted LED screen lifetimes ranging from 20,000 to 100,000 picture will look on an LED screen. The PC world extensively uses DVI hours are only meaningful if determined at the actual drive current (Digital Visual Interface – transports material between devices through used under real display conditions. a refining process), which is overwhelmingly popular (90%) in LED installations, and also becoming a standard for A/V. Video Processing A standard video signal cannot be directly displayed on an LED screen Electrical Usage and Maintenance without first processing it electronically. Many sophisticated electronics LED screen design and practices also include proper electrical adoptions now exits for image processing. Video images are comprised of a and maintenance. It is important to have the appropriate circuit number of horizontally scanned lines, but not all appearing on a screen breakers to handle in-rush prior to switch-on, overloads, and earth simultaneously. In the first 1/60th of a second (1/50th for PAL), the odd leakage. Finally when an LED screens needs a repair, you need to ensure lines are shown, followed by the even lines in the next 1/60th of a that right spare parts are available and that vendor service is adequate. second, forming a complete frame. This familiar interlaced display is When installed and configured properly, LED displays can be vibrant also used for television. For non-broadcast video signals, the signal is and exciting. By following these guidelines mentioned here, users can first de-interlaced. maximize results and avoid common pitfalls. Resolving video that involves rapidly moving objects requires LED displays are already a huge market ranging from cell phones, sophisticated interpolation in real-time along with appropriate scaling cameras, car dashboard displays to large-format displays. Still we can to fit the output screen that normally has a different resolution from expect a gigantic increase of these products worldwide in the near the source. These manipulations, especially the scaling, requires term. In the not too distant a future, we may also see OLED TVs and powerful processing to generate vivid, artifact-free, and fast-moving laptop screens as manufacturers work to increase their longevity.

Optics Background Photometric plots The required output specification is usually called out in photometric units: either Luminous intensity (lumens/steradian or lm/sr also known Microstructured Optics for as candelas or cd) or Luminance (cd/m²). Sometimes Illuminance (lm/ LED Applications m²) is important (mostly for uniformity requirements). Luminance is usually a calculated conversion from Luminous intensity > Arthur Davis, Reflexite Optical Solutions Business (it can also be measured directly). Luminous intensity is a quantity commonly measured at photometric labs and a typical output from Abstract raytracing. The formula for converting from Luminous intensity to Optics for use with Light Emitting Diodes are described. Microstructured Luminance is: optics are available and customizable for a wide variety of applications. A few of these will be touched on. A methodology of designing these optics and the photometrics of the typical technology is overviewed.

where L is the Luminance, I is the Luminous Intensity, A is the visible Introduction v v vis area and θ is the Zenith Angle. With the increasing popularity of LED’s in lighting applications, there is a need for engineered photometric control. Given exacting output There are a variety of ways to represent the photometrics. Luminous requirements, it is unusual for a given supplier’s LED to produce the intensity and Luminance can be represented on similar plot types. correct emission profile. This can be remedied with the use of auxiliary Useful representations are 3Dmesh, contour, polar, rectangular and optics. Available classes of optics include refractive (continuous surface Söllner. and microstructured), reflective (continuous and facetted) and 3Dmesh is uncommon, but it can help as a method of visualizing what diffractive. Examples are shown in Figure 1. This paper will concentrate the photometric distribution is. It is done by plotting the Luminous on microprism refractive optics with some mention of reflectors. intensity (or Luminance) magnitude in three dimensional coordinates. A 3Dmesh plot is shown in Figure 2 for a Luxeon LXHL-MW1A [1].

Figure 1: Common types of optics.

The type of designs considered here are light energy directing designs or “nonimaging optics”. Some nonimaging design background will be outlined followed by mention of some specific LED commercial Figure 2: LED Solid geometry with 3Dmesh photometric “batwing” distribution (Luxeon LXHL-MW1A). applications. Design methodologies for microstructured refractive A contour plot is made by taking the hemisphere in which the light optics will then be explored and a photometric analysis of a sample distribution emits and plotting the magnitude proportional to assigned design will be presented. Finally a discussion on some salient issues colormap values on polar coordinates as shown in Figure 3. The rings of regarding LEDs and microstructured optics is offered. the plot are the zenith (or the angle from the polar axis) and the spokes are the azimuth (or the angle around the polar axis). See Figure 4 for the geometry of the coordinates.

Figure 3: Contour plots for the photometric distribution shown in Figure 2. Left: Luminous Intensity. Right: Dis- Figure 5: Polar Plots. Left: Luminous Intensity Right: Luminance. crete level Luminance. The data from a polar plot can be equivalently plotted on rectangular axes. The x-axis is the zenith angle and the y-axis is the photometric magnitude. Again multiple slices can be overlaid with different colors as shown in Figure 6.

Figure 6: Rectangular Plots. Left: Luminous Intensity Right: Luminance.

The Söllner plot is another rectangular plot that is common for specifying

Figure 4: Spherical Coordinates. office lighting. Its y-axis is instead the zenith angle and the x-axis is a logarithmic scale of the photometric magnitude. The angular range is When a specification is considered, (such as max allowed luminance typically truncated to the region of interest (instead of 0°-90° for outside of a given angular range), it is sometimes useful to assign a small example 45°-90°). This plot lends itself to quickly checking whether a number of discreet contour levels with the maximum magnitude being maximum specified photometric magnitude is exceeded within a given the limit for the first contour level. Then at the given zenith limit, the angular range. See Figure 7. specification is easily observed by whether the first color level “bleeds across” the specification ring or not. For example, consider a specification that calls for less than 1000cd/m² outside of the 65° zenith angle. If the contour level from 0-1000cd/m² represents red, and >1000cd/m² is blue, on observing the plot, the spec is met as long as no red falls outside the region of the 65° zenith ring. This can be more informative than a polar plot which only takes several slices through the distribution, because the slices may “miss” a region where the specification is exceeded. This may not be an issue if the photometrics are strictly specified by the polar slices, but it is still useful to know if light is spilling out of the specified region. Figure 7: Semi-Logarithmic Plots. Left: Luminous Intensity. Right: Söllner plot for Luminance. A polar plot takes slices through a contour plot for constant azimuthal values. The polar axis on this plot then represents the zenith coordinate, Finally, often the light distribution at a given plane is desired. This can and the radial axis is the photometric magnitude. Typically multiple slices be observed by setting a detector at the selected location and then are taken (for instance at 22.5 degree increments) with each slice collecting and plotting the flux per unit area there. This is an illuminance overlaying the plot as a different color. An example is shown in Figure 5. map (see Figure 8).

Figure 8: Illuminance maps. Left: Total flux per unit area at a collection plane. Right: Same distribution with rays outside 10° excluded from the flux contribution; this is an approximation to what the eye might perceive.

A way to roughly simulate what the eye would perceive is to suppress the flux contribution of all rays that exceed a 10° observable cone. This is an approximation because the area of the flux distribution will usually be greater than the eye’s iris aperture. Furthermore, the foveal cone of Figure 9: Example Söllner plot including RP-1 specifications. The dashed cyan lines outline the preferred maximum average luminance values and the solid black lines outline the absolute maximum average luminance val- the eye does not subtend an angle that large. It is typically necessary to ues. The sample data sets are: Red: within the preferred range, Green: exceeds the preferred range, but still consider at least a 10° cone however because reducing the angle, for within the absolute allowed range, Blue: out of spec. example to 2°, greatly reduces the amount of rays in the flux distribution and produces a noisy illuminance map. Ideally however, visual simulation The maintained minimum Luminous Intensity for LED traffic signals are is achieved through a photorealistic renderer; see for example: excerpted from Table 1, section 4 of Reference 8 and reproduced in Radiance[2] (which powers Adeline [3]), POV-Ray [4], Photopia [5] and Table 2. To better grasp the numbers, a contour distribution can be Lightscape [6]. This topic is too vast to examine here, but it should be created by bilinearly interpolating the table values into a large polar recognized that many rendering software packages exist and one needs array. This is demonstrated for the red signal in Figure 10. to select a package that is sophisticated enough to import a source file representing the LED distribution. Horiz. Vertical Angle Angle Left & 8-inch Signal 12-inch Signal Specifications Down Right Red Yellow Green Red Yellow Green After generating the photometric analysis for a given design, the next task is to compare it to the requisite specification. Common lighting 2.5 133 617 267 339 1571 678 specifications include RP-1 [7] (for office lighting) and ITE [8] (for 7.5 97 449 194 251 1159 501 traffic signals). 12.5 57 262 113 141 655 283 2.5 17.5 25 112 48 77 355 154 The recommended preferred and maximum average luminances for 2.5 101 468 202 226 1047 452 office lighting are excerpted from Table 7-1 of Reference 7 and given 7.5 89 411 178 202 935 404 below in Table 1. For comparing photometric data to RP-1, it is useful to 12.5 65 299 129 145 673 291 overlay the specification boundaries on a Söllner plot of the data. To 17.5 41 187 81 89 411 178 satisfy the specification, the data should not exceed the luminance 22.5 18 84 37 38 178 77 boundary at the given angle (see Figure 9). 7.5 27.5 10 47 20 16 75 32 2.5 37 168 73 50 234 101 Degrees from vertical Preferred maximum Maximum average 7.5 32 150 65 48 224 97 luminance luminance 12.5 28 131 57 44 206 89 55° 850 cd/m² none specified 17.5 20 94 41 34 159 69 65° 350 cd/m² 850 cd/m² 22.5 12 56 25 22 103 44 12.5 75° 175 cd/m² 350 cd/m² 27.5 9 37 16 16 75 32 2.5 16 75 32 22 103 44 85° 175 cd/m² 175 cd/m² ≥ 7.5 14 65 28 22 103 44 Table 1: RP-1 specification. 12.5 10 47 20 22 103 44 17.5 9 37 16 22 103 44 22.5 6 28 12 20 94 41 17.5 27.5 4 19 9 16 75 32

Table 2: ITE specification [8] for minimum Luminous Intensity values for LED traffic signals.

www.led-professional.com LED professional Review | Sept/Oct 2008 | page 20

Figure 10: Bilinearly interpolated contour plots for the minimum luminous intensity distribution for red LED traf- Figure 11: Arbitrary distribution for sample specification. Left: Zenith increment scale is 10°; also shown are the fic signals according to the ITE specification8. The zenith increment scale (i.e. the grid formed by the circles) is 5°. polar slices: Red:22.5°, Green:45°, Blue:67.5°. Right: Zenith increment scale is 5°; also shown are the horizontal Note that the contour peak values are different although both plots are scaled to the same colormap. Left: 8” slices at the following declination angles: Orange:2.5°, Yellow:7.5°, Green:12.5°, Purple:17.5°. signal. Right: 12” signal.

For the purpose of a sample analysis an arbitrary specification is created. To compare the distributions, one can just look at the luminous This distribution has a minimum peak intensity of 6.0candelas at six intensities of the design and the specification side by side. But since the distinct propagation directions as shown in Figure 11. The specification actual specification is called out in terms of polar and rectangular slices shall require that the minimum luminous intensity exceed that of the through the photometric distribution, it is most informative to instead distribution everywhere along the following polar slices: 22.5°, 45°, slice both the data array and the specification distribution and then 67.5° and 90° (similar to an office lighting specification). In addition, it overlay the resulting plot lines on the same chart. The polar slice data should exceed the distribution at every horizontal angle at the following comparison for the sample Luxeon LED is shown in the top row of Figure declinations: 2.5°, 7.5°, 12.5° and 17.5° (similar to a traffic signal spec.). 12 and the rectangular slices in the bottom row. It is apparent that the specifications are not met. Auxiliary optics can be designed to shape the distribution to meet the specification. The design for an optical element to accomplish this will be consider in Section Design

Figure 12: Top Row: Polar slices through the four specified azimuth angles. The black line in each graph represents the minimum intensity requirement (each is the result of taking a data slice though the specified angle of the specification distribution). The colored lines are plots of the data through the output distribution. Since the colored lines dip below the peaks of the black lines, the specification is not satisfied. Bottom Row: Rectangular/ horizontal slices at the four declination angles. As before, the black lines are the minimum required and the colored lines are the output results. Again, the specification is not satisfied.

Applications is also wasted energy). Consequently a well engineered “side lobe” is required. The use of microstructured lenses and prism arrays makes all Frontlighting and backlighting this possible. Enormously popular applications of microstructured optics are in backlights and frontlights (see Figure 13). These are flat panel type waveguides used in LCD monitors, cell phones, pagers, PDAs, handheld games and other displays. Especially when using LEDs, fine photometric control of the light distribution is necessary in order to present the observer with a uniform observation field. This is usually accomplished with some form of waveguide preconditioning or integrating optic. This optic serves to illuminate the edge of the flat panel waveguide with a uniform input. Once inside this waveguide, microstructure is heavily relied upon. By either printed or etched dots or rows or patches of prisms, the light is selectively transferred out of the waveguide so that the whole panel appears uniformly illuminated. Still afterwards, more microstructure is important in providing the observer with the requisite screen brightness. Often a prism film is used to increase the gain and efficiency while recycling light that would otherwise be wasted.

Figure 14: A traffic signal sample design that uses a high intensity LED module in rear-projection fashion.

Flashlights In an LED flashlight application, multiple LEDs may be used as the light source. This array is larger than a single typical flashlight bulb and for that reason it may be too unwieldy to create an effective reflector. Light collection and collimation can be improved by using a Fresnel lens on the output of the flashlight instead.

Office lighting Figure 13: Schematic of a typical backlight. In an LED application, the fluorescent lamp would be replaced with Although the author currently does not know of any office lighting multiple LEDs and an integrator optic. fixture yet that implements LEDs, it is probably not unreasonable given the current state of the technology to design an RP-1 compliant direct/ Traffic and rail signaling indirect fluorescent/LED hybrid. The downlight requirements for direct/ A popular national trend is to replace incandescent traffic control indirect illumination are not extreme so using optically conditioned signals with LED signals. There is a well defined standard for the high intensity LEDs is plausible. At the same time, all uplight contributions luminous intensity performance of these signals (see ITE specification can be provided by the fluorescent (instead of splitting it between in Table 2 and Figure 10). Accordingly, any light that is directed upwards uplight and downlight) so a fixture that normally uses two fluorescents is a waste of power and can be redirected with the help of microstructured may potentially be replaced by a fixture that uses a single fluorescent in optics. Additionally, new designs with microstructured optics eliminate combination with LEDs. The color rendering of such a hybrid however the unpleasant “discreetness” of large LED arrays by using higher may be unusual. intensity modules in a rear projection fashion (a example is shown in Figure 14). Custom lens elements implementing microstructured optics Video Projection are ideal for redirecting these sources to a pleasantly uniform ITE Potential for microstructured optics may yet exist in video projection specification. with the advent of newer LEDs. Traditionally plastic optics cannot be A similar application is in the Rail signal industry. Rail signals are applied as source conditioners (i.e. field lenses and lens arrays) because different from traffic signals in that they are mostly viewed in-plane the close proximity of the plastic to the extremely high temperatures of (instead of significantly declined from the optical axis). They also have the incandescent source damages the optics. It would be highly desirable the unique requirement of being visible from far away and all throughout to use plastic optics in this application over glass because of plastic’s the subtended range a train operator perceives while traveling past the lightweight. Should LED brightness become sufficient for video projector signal. Additionally, it is undesirable for train operators on the opposite applications, their inherent high efficiency and low thermal output side of the sign to have visibility of the light (light directed in this region (when compared to a comparable incandescent source) could make the use of plastic optics for lamp conditioning much more reasonable.

Additionally, there is the option of molding in a Moth-eye structure [9] Sequential raytrace packages excel at quickly raytracing and optimizing for improved efficiency. Aside from the incremental tooling cost, process an optical layout. The great speed of the software comes from the cycle time and manufacturing unit cost are marginally increased assumption that light will always travel to incidence on the next optical compared to a similar part without Moth-eye. In comparison to thin film surface defined in the database. This is perfectly acceptable for imaging coatings, the performance and value is enhanced considerably [10]. systems. Also adding to the power of most sequential raytrace software is the inclusion of global optimization algorithms in the software. By Miscellaneous defining a figure of merit (such as spot size) and some system variables Of course the potential applications for LEDs are innumerable. For any (such as radii of curvature), the global optimization routines can modify of those applications where fine photometric control is required, or the system automatically until optimum performance is achieved. Popular simply a different distribution from the manufacturer supplied LED is software in this category is OSLO [16], Zemax [17] and CodeV [18]. preferred, microstructured optics may be used in conjunction. More For nonimaging systems and for stray light analysis of imaging systems, examples would include sign illumination in which optics can more where light does not necessarily travel sequentially, a non-sequential evenly distribute the bright LED spots over the face of the sign, artistic raytrace package is required. In a non-sequential raytrace, light is lighting in hallways, movie theatres and galleries, marker lighting for permitted to reverse direction and become incident on optics it has roads and runways in which a cylindrical upward inclined distribution is already traversed, it can get trapped between surfaces, or it can skip desired, machine vision, automatic teller bank machines, pedestrian the next optic in sequence altogether. Because of this generality, non- signs, automotive interiors such as for the instrument panels and for sequential raytracers trace rays much slower. This tends to make the dome lights, automotive exteriors such as for the brake and signal automatic global optimization more difficult to implement. Popular lights, street lighting, ring lights etc. software in this category is ASAP [19], TracePro [16] and LightTools [18]. ASAP has a linear optimizer built in which can be readily applied in the Design design process. TracePro and LightTools have none, but the underlying A concise description of Illumination System design is given in Reference macro languages are substantial enough to write one from scratch. 11. To summarize, the process involves: building the fixed geometry, In order for a raytrace simulation to be as close to reality as possible, it accurately modeling the source, determining the requirements, is best to have the actual source photometrically characterized. This can evaluating simple initial concepts, optimizing a selected concept, be done by accurately measuring the source in a photometric lab. A verifying the performance given geometric deviations, tolerancing, typical photometric report will include the photometrically accurate far prototyping and finally testing [11]. The sample design presented in this field luminous intensity values. From this data, careful calculation can section loosely follows these tenants. No geometric deviation or result in the definition of a source model that is approximately apodized tolerancing is performed however and as of this writing the photometrics in accordance with the real source. It is most convenient however to use for the sample design presented have not yet been verified by testing. a software package such as ProSource [20] coupled with photometric measurements supplied by Radiant Imaging [20]. This allows the designer Optical modeling tools to generate accurate ray distributions via a nice user interface. There are numerous ways to create a model of an optical layout. Tools of the trade commonly consist of a mechanical CAD package, a Other useful software includes scientific data visualization software sequential raytrace package with global optimization, a non-sequential (IDL [21], PV-Wave [22], Origin [23]) and capable programming raytrace package and radiometry ray generation software. languages (C/C++, Perl, Fortran and many others). Data visualization is an important tool in combination with raytrace packages because the One of the most straightforward ways of setting up a layout for data analysis portions of raytrace software are comparatively limited raytracing is to draw the setup in a mechanical CAD package (SolidWorks for purposes of calculation and publication. Programming language [12], ProEngineer [13], Rhinoceros [14], AutoCAD [15] to name just functionality adds similar capability, and also can be used to augment several). Once completed most CAD programs can export to a format the raytracing software functionality. that can be imported by most raytrace packages. This methodology can be tedious for design optimization because each decided change in Choosing an LED geometry requires redrawing the system, re-exporting it from the CAD Often, a customer seeking a tailored photometric output will already package and re-importing it to the raytrace software. The process can have an LED specified. It is most convenient for the designer if the LED be streamlined somewhat by using parametric software such as is common enough to have been characterized and published in the SolidWorks or ProEngineer. Parametric CAD software defines its Radiant Imaging Source database [20]. Any and all photometric data geometry according to variable constraints. By judiciously constraining that can be supplied is useful in the design. It is also appreciated when the model, the entire geometry can be quickly regenerated just by a supplier of LEDs has gone through the effort of characterizing their changing the value of a variable. LED sources and provide the data as requested.

If the design engineer has the liberty to choose an LED from the start, a significant factor in choosing a supplier is the free availability of characterization data (for example, downloadable photometric characteristics and published mechanical CAD drawings). This reduces the initial steps of acquiring varied samples of LED’s for measuring their characteristics. With photometric data readily available, a design feasibility study, (and indeed a full design) can be completed without having to receive a physical sample.

LED source modeling Before considering how to shape the design photometrics with optics, it is paramount that an excellent source representation be used. Figure 16: LED with absorbing shell (blue) just surrounding the transmissive dome (red). Approximations and inaccuracies will at best transfer linearly through the system effecting error in the results. More likely though, errors will be compounded through the ray propagation yielding extra unrealistic results and incorrect flux magnitudes. (Concerning source modeling, see for example, References: 24, 25, 26 and 27.). The LED source model can be simulated either by geometric detail or by ray generation based on accurate photometric measurements of the LED [24, 27]. Raytraces based on a characterized source model generally produce photometrics that are more accurate than a strictly geometric source model. However, if the source is to be used in a system where light may be recycled and impinge back on itself, it is essential to have its geometry included. The following merges the CAD geometry for a Figure 17: Left: Absorbing shell for collecting rays outside of the LED. The spot cloud shown is the original ball of Luxeon 1-Watt white LED (LXHL-MW1A) [1] with the Radiant Imaging rays that would lie inside the LED geometry. The LED geometry has been removed so these rays can be free-space generated rayfile for this LED. propagated to the collection shell. Right: Absorbing shell with traced rays. Rays that miss the shell are presumed to be from stray light and are dropped. Figure 15 shows several views of the CAD geometry for the LED with the rayspots overlaid. The origin for both the CAD and the ray Next the rays are propagated (through free space) to be absorbed at the representations are coincident and is located at the center where the shell as shown in Figure 17. Note that some spurious rays are not glass envelope meets the base. ProSource creates rays at arbitrary coincident on the shell. In reality, this is not plausible because all emitted starting positions. As shown in Figure 15 all of the rayspots originate light should only come through the transmissive window of the LED. inside the boundary definitions of the LED. If the LED geometry is to be Most likely this is caused by ProSource reconstructing ambient scattered included in the raytrace, this is not where the initial ray coordinates light. These rays are a small fraction of the total flux (less than 0.5%) belong. ProSource generates the rays from the perspective of an and are deliberately dropped and ignored in further raytraces. The rays observation plane after the rays have already left the LED. Putting the that are captured at the shell are transferred to a new rayfile to be used rays back behind the glass dome and propagating them through again for all future raytraces. The CAD geometry for the LED is added back in is not correct. To get the rays “out” then, they should be free-space and the location of the rays are plotted in Figure 18. propagated to an exterior dummy surface. This is accomplished with an absorbing shell located to surround the transmissive portion of the LED (Figure 16). The LED geometry is removed, and the rays are inserted (Figure 17).

Figure 15: Views of the LED with the originating source rays overlaid. The red dome feature is the glass envelope Figure 18: LED with externally captured rays. The absorbing shell is removed and the new ray positions are saved and the black clipped cylindrical feature is the opaque base. for future raytraces.

Reflector design Refractive microstructure design The science of designing a reflector to maximally concentrate light from • Layout a source emitter is well researched (many useful publications are The approach taken for designing a refractive microstructure is highly compiled in Reference 28). One type of reflector which can maximally dependant on the application. For designing and manufacturing redirect light output from an input plane is called a Compound Parabolic components in a well developed field, the design can practically be Concentrator or CPC [29, 30, 31, 32]. automated and therefore completed at little to no cost. An example is A schematic of a CPC is shown in Figure 19. As per the variables defined the design of a collimating Fresnel Lens. Since generalized software has in that Figure, the following are some useful formulae for CPC design been written to create facet geometry according to a lens prescription, [29, 30]: the lens design is practically a commodity. The most significant costs are for tooling and manufacturing. A stringently specified project will be much more complex. As an example, take the arbitrary specification fabricated in Section Specifications, Figure 11. Accordingly, the intent of this section shall be By choosing a smaller value for θmax, a higher degree of collimation is to take the on axis output from the LXHL-MW1A seated in the CPC and achieved. However, this also very quickly increases the requisite length create six radiating off axis spokes. To begin, it is necessary to choose (L) of the CPC. This can be traded off with not extending the CPC length an optical structure of approximate proper form. This will give the out to L. Hinterberger [29] notes that truncating the reflector at 2/3L optimization process its necessary starting point. In this case we will does not decrease the aperture diameter very much. Although, in his select a six-sided pyramidal structure; each facet of the pyramid will application, the CPC is used to collect light onto the focal plane (instead direct light toward one of the spoke directions [34]. of collimating a source placed at the focal plane). When truncating the CPC length for source collimation, careful attention should be paid to Early in the design process, it is substantially important to consider its effect on the output half-angle. tooling and manufacturing. Given the unlimited design flexibility within the scope of design software, it is easy to create a design that cannot be tooled or is prohibitively expensive to manufacture. It is frustrating to put significant effort in developing an optimal design to only later find out that it cannot be made. Sometimes it is possible to alter the design to make the part realizable, but it will inevitably be at the cost of performance. Usually it is best to include tooling and manufacturing constraints within the optimization process. Depending on the tooling process used, different constraints may exist. In this case, the tooling is made by diamond ruling. We can form a six sided pyramidal structure by engraving crosscuts inclined to each other by 60° (see Figure 20). This will not create standalone pyramids; sub- prism features will also be formed. Understanding these features and including them in the design process is essential. To accomplish this, it is useful to make a three-dimensional representation of the solid Figure 19: A Compound Parabolic Concentrator (CPC). Variables: d: Source plane diameter; d2: Exit aperture geometry in CAD software (see Figure 20). diameter; L: CPC length; θmax: Output half-angle.

It can be shown that if a CPC reflector is converted into a solid dielectric with an index of refraction n satisfying: 2 ≥ n ≥ √2 (a range of indices covered by common materials) then total internal reflection is supported at the CPC walls.� This method can have significant cost benefit by eliminating the need to add a reflective coat. Furthermore, an engineered microstructured surface can be incorporated into the output surface for additional photometric control. In this way, both the light collector and a refractive control structure can be formed by a single injection mold.

At least one company capable of supplying CPC reflectors on a custom Figure 20: Left: The tool is engraved at three separate cut directions (red, green and blue) to form a repeating 6‑sided prism structure (shaded regions). Right: The solid geometry formed by ruling the pattern shown at the left. manufactured basis is Opti-Forms [33].

www.led-professional.com LED professional Review | Sept/Oct 2008 | page 26

On considering pitch, it is usually not so important to design the If after experimenting with a few manual optimization iterations, it microstructure to the exact dimensions of the finished product. When becomes apparent that too many factors are interacting to seek a the far field luminous intensity distribution is the specification of resolution by hand, then an automated optimizer should be applied. concern, often 10 prisms will do almost as good a job as 1000. The This is much more time consuming to set up initially, but it is far more requirements are that the facet angles be maintained and the ratio of efficient at characterizing large numbers of possible geometries then a geometry feature dimensions remain fixed. The size of the microprisms manual process. can usually be chosen to satisfy the aesthetic requirements (whether The first step in automated optimization is to define an error function the features should be visible to the unaided eye or not). Also tooling (or figure of merit). The error function is some way of quantitatively may be considered in selecting the pitch. Larger feature sizes decrease defining how good the system performance is with a single number. As the time it takes to make the tooling master. Especially fine structure the error function approaches zero, the system performance improves. can approach the limit of tooling capability in addition to degrading The design of an appropriate function can be complex. In this case, and performance as the ratio of feature size to tooling error (burrs, radiused in many other cases where just the luminous intensity distribution is of edges etc.) decreases. concern, it can be simple. The error function can be defined as the Furthermore, it is beneficial to exploit large facet size in the computer difference between the total flux of the LED and the amount of flux model itself. Generating numerous quantities of prisms in the model contained within the desired propagation spokes. If all flux was requires the computer simulation to solve a plethora of ray-boundary contained within the spokes, then the error function would be zero and intersections. This is time consuming even on the currently fastest an optimum design would be signified. In this example, there is no available computer hardware. Design iterations can proceed much more restriction on cutoff angle, but if a specification were to call for an efficiently if larger size prisms are considered. An additional trick which explicit cutoff, an additional factor should be added to the error function can reduce the number of prisms stored in the raytrace database is to to strongly punish the system merit until the cutoff requirement is generate only several rows of prisms and then to bound them on either satisfied. side by perfectly reflecting mirror planes. It is important that the planes The next step is to figure out what parameters of the system may be be located such that symmetry of the facet structure is preserved (that allowed to vary in order that an optimal geometry will be obtained. is, in unfolding the prisms about the mirror planes, the microstructure Typical parameters are facet slope angles, draft incline angles and pitch should appear uninterrupted). This technique is not always possible and to height ratios. To follow, much geometric calculation must be done to it is inaccurate at the boundaries, but for design purposes, it saves a lot fully define the geometry based on these parameters. Care must be of time. After design is complete, or during the last few iterations, the taken to neither over-constrain nor under-constrain the geometric pitch may be reduced to actual size and the full volume of prisms definitions. With each optimization loop, the geometric definitions will generated (removing the perfect mirror planes) to verify the be referenced procedurally. The procedure call will only provide the photometrics. This may not always be possible however when particularly designated parameter variables from which the entire geometric small prisms are applied over a very large area; such great numbers of representation must be created. It is also useful to include some reality geometric definitions may exceed the capabilities of the software or checking functionality to weight the algorithm against driving the take eons to raytrace. system towards a physically unrealizable geometry. With some of the new features being implemented in raytrace software, Before hitting the “go” button, there will be a number of remaining the issue of pitch and number of prisms affecting raytrace time can details to wrap up depending on the application. Of course the become mute. If the prism structure is simple enough, the virtual underlying flow control of the optimization process must be microstructure ability of RepTile in TracePro [16] and 3D‑Textures in implemented. Some software includes built-in optimization routines. LightTools [18] can apply vast numbers of prisms to a surface with the For others, it may be necessary (or preferable) to write one from scratch definition of a single facet period. (some of the common software available is briefly discussed in Section Optical modeling tools). Also, optimization time should be considered. • Optimization A balance must be struck between accuracy and iteration time by deftly Once a the basic geometric form is developed, a process of optimization manipulating the following parameters: may commence. This can range anywhere from completely manual to nearly fully automated. In a manual optimization, each iteration would • number of rays to trace involve regenerating or reprogramming the facet geometry based on • ray splitting and/or ray scattering intuition and on how previous designs performed. With moderately • ray termination stringent photometric specifications, and a solid grasp on how the optical facets interact with the light input, a manual optimization is • geometric sampling resolution completely reasonable and may be the quickest way to achieve a final • feature periodicity (when not using virtual microstructure) design.

If all goes well, then a solution may be found. Keeping track of all the Results geometries and the results of their merit function is paramount when a The photometric results are analyzed as described in Section Photometric large number of iterations are traversed since the terminating solution plots (see Figure 21). These results are compared to the specification of (or more likely, the current solution being analyzed when you hit the Figure 11 as described in Section Specifications (see Figure 22). The “cancel” button) may not be the best system. This may be accomplished photometric requirements exceed the specified minimum values at all with supplementary software or scripting programs designed to analyze angles and therefore satisfy the specification. the wealth of raytrace output files and compile the results into a The predominant feature of generating six off axis “spokes” is achieved. convenient table. The table should include enough information to As shown in Figure 21 however, there is an added on-axis lobe to the recreate the system. It is also convenient to sort the results in terms of distribution. This lobe is not strictly forbidden by our specification. As the error function and to include other analytical numbers such as such, it should not be a problem. Admittedly though, it is a source of output flux, efficiency and so forth. inefficiency, robbing light from the specified lobes. Also, although the If an acceptable solution is not found, the optimization process starts on axis lobe may not be prohibited in a customer’s original specification, over again. This can include anything from redefining the original a dialog must be opened to communicate the feature and prevent geometry paradigm, to providing more (or less or different) degrees of surprise. Without having realized it before, the customer may find that freedom in the way the algorithm modifies the geometry or to simply some photometric quirks actually are unacceptable and should be altering the initial conditions of optimization. added to the specification. Still, given time and cost constraints it must be decided if any such peculiar photometric incarnations are acceptable. This also needs to be balanced in conjunction with the realization that further complicated design may yield microstructure features that prohibitively increase the tooling cost.

Figure 21: The various photometric plots for the output of the optimized sample design.

Figure 22: Comparison of the sample design photometrics to the defined specifications. The specifications are satisfied.

Discussion Source obscuration The non-uniform illuminance pattern of the LED near field causes the LED vs. fluorescent significant complaint of source visibility. This generally occurs in In designing microstructured optics for traditional lighting applications applications where an LED is closely spaced to an optic. On the other that use fluorescent sources, most of the methodology previously side of this optic is the requirement for no source visibility with highly discussed for LEDs is directly applicable. A typical difference is the wide uniform illumination. This is a fairly tall order. Commonly, a high angle availability of photometric characterization for LEDs, which are highly diffuser is used. A chief problem is that with said diffuser placed closely accurate. Whereas such characterization files for fluorescent lamps are to an LED, its near field illuminance pattern is projected onto the diffuser not freely published and are more expensive to generate. Thereby it is and is consequently immediately visible and objectionable. A standard more common to use a less accurate geometrical model for fluorescent solution is to space the source as far away from the output optic as lamps. Furthermore, the emission cone of an LED is much more well possible and use multiple diffusive layers maximally spaced. This is confined then a cylindrical tube. This makes it easier to control the typically unwanted because it uses a large volume and is inefficient. output distribution at a high efficiency since the LEDs can be spaced Under the constraints of small LED to lens distances, the solution to such that there is no or minimal overlap of the light distribution on the uniform output lies in light recycling. Injecting the LED source optic between neighboring sources. Conversely, this makes achieving a distribution into a tailored waveguide can be used to pre-condition the visual uniform field more difficult. light into a uniform field. Waveguide extraction however can be an inefficient process. Some solutions to this problem are currently being investigated at Reflexite Display Optics with a strategic partnership for manufacturing technology that is much more highly efficient and can effectively obscure the source over a short distance.

Rigorous design vs. “plug and play” Environmental conditions The rigorous computer design method set forth is the recommend Just as attention must be paid to the environmental conditions under design paradigm, however, it is still quite popular to design with the which an LED will be subjected, so too should concern be paid for the “plug and play” approach. That is, buy some LEDs, maybe grab a lens, microstructured optic. Typical materials are Acrylic, Polycarbonate or install it in the system, turn it on and see if it works. For loosely defined Polystyrene. Zeonor [35], Topas [36] and Arton [37] may also be used photometric requirements, indeed this might be the preferred design for very high temperature applications. The environmental conditions method because it is cheap and fast. If during the course of “plug and to consider are mechanical stress for durability, temperature for heat play” design one finds that trying multitudes of LEDs with random distortion, humidity for water absorption and ultraviolet light exposure optics just is not working, a custom design would probably help. It may for yellowing. Also material density may be important for weight even be preferred to seek a customized design even when a plug and considerations. play solution is available; due to the precise photometric control that can be exercised with a microstructured optic, a photometric output Manufacturing may be sought that directs all light where it needs to go with very little Regarding manufacturing process, trade-offs, as usual, can be stray energy outside of the requisite minima. This ensures maximal light considered between cost and performance. The primary mechanisms in utilization and promotes power conservation. these trade offs for microstructured optics are fidelity, internal stress (which manifests as birefringence), precision and aspect ratio. A Inter-Compatibility compression molding38 process has a long cycle time as compared to For any project in which an LED source is to be used, the designer should an injection molding process. The faster cycle times of an injection also take into concern obsolescence. As manufacturers refine their molding process can reduce unit manufacturing cost. This comes at the LEDs, it is not always a simple matter to unplug the old model and “drop expense of prism fidelity, added birefringence and lower precision. Flow in” a new one. For example, given that the trend is to increase brightness, front freezing occurs in the plastic as the grooves are filled out and for specifications that indicate a maximum luminance for a given cutoff instead of sharp features being formed, a rounding out is caused (see angle, dropping in a brighter LED may exceed acceptable cutoff. Also, Figure 23) [39]. As a result, the injection molded part is less efficient for a tightly toleranced layout in which the emission profile of the LED than the compression molded part. For a cost sensitive or high volume is precisely controlled, dropping in a new LED would be flawed if the application in which efficiency, low birefringence and precision are not new LED had a slightly different emission profile. For this reason, it is as important, this may be entirely acceptable. Of note, there are also preferable that LED manufacturers carefully maintain LED emission hybrid processes known as Injection Compression Molding (coining) profiles to be constant through LED generations. and High Precision Molding (HPM) [9] which enable cycle times close to that of injection molding while achieving fidelity, birefringence and At the discretion of the engineer based on the project requirements, precision close to that of compression molding. system dependence to a specific LED can be minimized by loosening up the tolerances (usually at the expense of some system performance). Then when a new LED is installed it may be possible to bring it into specified performance by slightly adjusting the optic to LED distance. That is to say, adjustment of the emission profile may be tweaked by “focusing” the lens to different locations based on the LED. Allowing for such mechanical adjustments has the added benefit for optics that might be used simultaneously for several different color LEDs or several different manufacturer’s LEDs (or both). Each unique LED may have a Figure 23: Side by side comparison of fidelity when molding microprism facets with: Left: compression molding distinct “source to optic” distance. The final output profile for each optic and Right: injection molding. however can be very similar. The alternative is to change the optical element surface structure per LED type. Each LED type would require its Aesthetics own redesigned optic and new molding tools would have to be made. Aesthetic expectations for a microstructured optical design can Additional molding tools would have to be made later as newer sometimes cause trouble. Often a customer wants a certain look and generation LEDs replaced older ones. “feel” to the finished product. This is not easily communicated through an input specification. Furthermore, it is not easily comprehensible When considering defocusing an optical element as a means of throughout the design process to even tell what the visual impact of a compensation over creating a new lens, it should be noted that one also finished product will be until prototypes have been made. Prototyping needs to balance the added cost of altering the mechanics of a system microstructured optics is a significant expenditure so it is best to get it to accommodate a different lens location. If expensive redesign is right the first time. In many instances, the photometric performance of necessary to defocus an optical element, it may be more economical to a particular design may well meet or exceed the photometric make a new optic that functions in the existing system architecture. specifications, but if the aesthetic quality is not sufficiently pleasing to the customer then redesign or project termination will still occur.

Often, it is the presence of a glaring view of the source that is The way in which performance is evaluated varies from industry to objectionable. Such a problem may not be obvious in a standard industry. Some of the standard ways in office lighting and signage were raytrace. The engineer is then charged with using intuition to scrutinize presented and are broadly useful over a great many applications. The the geometric model and the raytrace to see if any glare paths are specifics of evaluation however has to be determined on a case by case apparent. Ideally, aesthetics may be evaluated via very high end basis. photorealistic rendering (this topic is touched on at the end of Section The design approach presented within was followed for a contrived Photometric plots) before going to prototype. Otherwise some form of example. The procedure for any other real world application will approximation must be used. For example, instead of tooling and naturally be different, but the underlying principles are valid for most molding the part, a cheaper prototype can be made by machining the situations. microstructure directly into plastic. Finally there is a wealth of topics worth pondering when considering Summary merging LEDs with microstructured optics. Some of the issues discussed were: LEDs versus fluorescent sources, source obscuration, rigorous A wide range of topics have been touched on through this discourse; design vs. “plug and play”, inter-compatibility of different LEDs for a each fully deserving of its own treatise. The applications for LEDs and given design, environmental conditions, manufacturing trade-offs for microstructures in their own rites are vast. The combination of the two cost with performance and finally aesthetic expectations. is a logical symbiosis.

Acknowledgements: Thanks to Dave Jenkins and Doug Kreysar of Radiant Imaging20 for explaining concepts regarding how ProSource software generates rays. Thanks also to Michael Foley, Bryan Parks, Joe Lupone and Donna Desmarais for input regarding content and corrections. References: * Entries marked with this symbol have reprints available in Reference 28. 1 Lumileds Lighting, LLC, 370 West Trimble Road, San Jose, California, 95131, Tel: 877-298-9455, Fax: 408-435-6855, Email: [email protected], http://www.lumileds.com 2 Radiance Synthetic Imaging System, http://radsite.lbl.gov/radiance/ 3 Adeline, Fraunhofer-Institut für Bauphysik, http://www.ibp.fhg.de/wt/adeline/ 4 POV-Ray, http://www.povray.org/ 5 Photopia, Lighting Technologies Inc., 1630 Welton St., Suite 400, Denver, Colorado 80202, Tel: 720-891-0030, Fax: 720-891-0031, Email: info@lighting‑technologies.com, http://www.lighting‑technologies.com/ 6 Lightscape, http://www.lightscape.com/ 7 Illuminating Engineering Society of North America, “IESNA RP-1-1993”, http://www.iesna.org 8 Institute of Transportation Engineers, “Vehicle Traffic Control Signal Heads (VTCSH)” and “VTCSH Part 2: LED Vehicle Signal Modules (Interim)”, http://www.ite.org/standards/led.htm 9 M.F. Foley, “53.4: Microstructured Plastic Optics for Display Applications”, SID Digest, Volume XXX, pp. 1106-1109, (May 1999). 10 http://www.display-optics.com/pdf/moth-eye.PDF 11 David Jenkins, and Mark Kaminski, “Using Computers to Design Nonimaging Illumination Systems”, SPIE Vol. 3130, pp. 196-203 (29-30 July 1997) 12 SolidWorks Corporate Headquarters and Northeast Region, 300 Baker Avenue, Concord, MA 01742, Tel: (978) 371-5000, Fax: (978) 371-5088, Email: [email protected], http://www.solidworks.com 13 PTC, 140 Kendrick Street, Needham, MA 02494, Tel: 781-370-5000, Fax: 781-370-6000, http://www.ptc.com 14 Robert McNeel & Associates, 3670 Woodland Park Avenue North, Seattle WA 98103, Tel: 206-545-7000, Fax: 206-545-7321, http://www.rhino3d.com 15 Autodesk, Inc., 111 McInnis Parkway, San Rafael, CA 94903, Tel: 415-507-5000, Fax: 415-507-5100, http://usa.autodesk.com 16 Lambda Research Corporation, 80 Taylor Street, P.O. Box 1400, Littleton, MA 01460-4400, Tel: 978-486-0766, Fax 978-486-0755, http://www.lambdares.com 17 Focus Software, Inc., P.O. Box 18228, Tucson, Arizona 85731, Tel: (520) 733-0130, Fax: (520) 733-0135, Email: [email protected], http://www.zemax.com 18 Optical Research Associates, 1800 West Park Drive, Westborough, MA 01581, Tel: 626-795-9101, Fax: 626-795-9102, http://www.opticalres.com 19 Breault Research Organization, Inc, 6400 East Grant Road, Suite 350, Tucson, Arizona 85715, Tel: 800-882-5085, Fax: 520-721-9630, Email: [email protected], http://www.breault.com 20 Radiant Imaging, Inc., 26425 NE Allen St., Suite 203, Duvall, Washington 98019, Tel: (425) 844-0152, Fax: (425) 844-0153, Email: [email protected], http://www.radimg.com 21 Research Systems, Inc., 4990 Pearl East Circle, Boulder 80301, Tel: 303-786-9900, Fax: 303-786-9909, Email: [email protected], http://www.ResearchSystems.com 22 Visual Numerics, Inc., 2000 Crow Canyon Place; Suite 270, San Ramon, CA 94583, Tel: 800-364-8880, Email: [email protected], http://www.vni.com 23 OriginLab Corporation, One Roundhouse Plaza, Northampton, MA 01060, Tel: 1-800-969-7720, Fax: 1-413-585-0126, http://www.originlab.com 24 Ronald F. Rykowski and C. Benjamin Wooley, “Source Modeling for Illumination Design”, Proc. SPIE Vol. 3130 (1997). 25 David R. Jenkins and Holger Mönch, “P-81: Source Imaging Goniometer Method of Light Source Characterization for Accurate Projection System Design”, 2000 Society for Information Display, Long Beach, CA (2000). 26 William J. Cassarly, David R. Jenkins, and Holger Mönch, “Accurate Illumination System Predictions Using Measured Spatial Luminance Distributions”, Proc. SPIE Vol. 4775 (2002). 27 John F. Van Derlofske, “Computer modeling of LED light pipe systems for uniform display illumination”, Proc. SPIE Vol. 4445 (2001). 28 Roland Winston, (Editor), Brian J. Thompson, (General Editor), Selected Papers on Nonimaging Optics, SPIE Milestone Series, Volume MS 106 (1995). 29 H. Hinterberger, and R. Winston, “Efficient Light Coupler for Threshold Čerenkov Counters”, Review of Scientific Instruments, Vol. 37(8), pp. 1094-1095, (Aug. 1966).* 30 H. Hinterberger, et. al., “The Design and Performance of a Gas Čerenkov Counter with Large Phase-Space Acceptance”, Review of Scientific Instruments, Vol. 41(3), pp. 413-418, (Mar. 1970). * 31 H. Hinterberger and R Winston, “Use of a Solid Light Funnel to Increase Phototube Aperture without Restricting Angular Acceptance”, Review of Scientific Instruments, Vol. 39(8), pp. 1217-1218, (Aug. 1968). * 32 Roland Winston, “Light Collection within the Framework of Geometrical Optics”, Journal of the Optical Society of America, Vol. 60(2), pp. 245-247, (Feb. 1970). * 33 Roland Winston, “Dielectric compound parabolic concentrators”, Applied Optics, Vol. 15(2), pp.291-292, (Feb. 1976). * 34 Opti-Forms, Inc., 42310 Winchester Rd, Temecula, CA 92590-4810, Tel: 909-296-1300, Fax: 909-296-1178, http://www.optiforms.com/ 35 The design herein presented was not carried through the actual processed described. Indeed, the “arbitrary” specifications were specifically tailored to match the performance of the chosen part. In reality, the chosen part was selected because it was a non-proprietary design with good aesthetics and the compression mold tooling was readily available. The method of design however is still valid and has been applied towards other optical elements with different specifications. 36 Zeon Chemicals L.P., 4111 Bells Lane, Louisville, KY 40211, Tel: 877-275-9366, Fax: 502-775-2055, Email: [email protected], http://www.zeonchemicals.com/ 37 Ticona, 90 Morris Avenue, Summit, New Jersey 07901, Tel: 908-598-4000, http://www.ticona‑us.com/ 38 JSR Corporation, http://www.jsr.co.jp/ 39 John R. Egger, “Manufacturing methods for large microstructured optical components for non-imaging applications”, SPIE Vol. 2600, pp. 28-33, (Dec. 1995) 40 Arthur Davis, Robert C. Bush, John C. Harvey and Michael F. Foley, “P-95: Fresnel Lenses in Rear Projection Displays”, SID 2001 Digest, Volume XXXII, pp. 934-937.

in nonimaging optics that cannot be solved with systems using linear or Review of SMS Design rotational symmetry. After overcoming complex obstacles in both programming and 3D surface creation, the SMS 3D method now has Methods and Real World matured and reached a level of stability that makes it an every day Applications design tool. Some of the results achieved are presented below. Up to this point, the SMS method has mostly dealt with two pairs of > Oliver Dross et al., Light Prescriptions Innovators LLC wavefronts to be coupled. The design results are two profiles or surfaces. The Simultaneous Multiple Surfaces design method (SMS), proprietary Many extensions of this method are being developed, some of which we technology of Light Prescription Innovators (LPI), was developed in the will presented in the last chapter of this article. early 1990’s as a two dimensional method. The first embodiments had either linear or rotational symmetry and found applications in SMS 2D Review photovoltaic concentrators, illumination optics and optical In SMS [1] [2] (simultaneous multiple surface) design procedures, the communications. SMS designed devices perform close to the optical prescription is stated as incoming and outgoing wavefronts. The thermodynamic limit and are compact and simple; features that are definition of these wavefronts is sometimes trivial (see example below) especially beneficial in applications with today’s high brightness LEDs. but in other cases highly complex: If the design goal is to meet a certain The method was extended to 3D “free form” geometries in 1999 that irradiance or intensity distribution, there is no deterministic relation perfectly couple two incoming with two outgoing wavefronts. SMS 3D that allows the derivation of exiting wavefronts that create said controls the light emitted by an extended light source much better than distribution. However, approximate methods may be applied [3]. single free form surface designs, while reaching very high efficiencies. In the following section we will review, at a simplified level, the design This has enabled the SMS method to be applied to automotive head methods that consist of coupling two pairs of wavefronts that create lamps, one of the toughest lighting tasks in any application, where high two generally free form profiles (SMS 2D) or surfaces (SMS 3D) that can efficiency and small size are required. This article will briefly review the be either refractive or reflective. The optical system may consist of any characteristics of both the 2D and 3D methods and will present novel number of surfaces that exist before, after or in between the SMS optical solutions that have been developed and manufactured to meet surfaces. These surfaces are either predefined by the application (e.g. real world problems. These include various ultra compact LED collimators, the dome shape of an LED light source or a fixed exit aperture profile) solar concentrators and highly efficient LED low and high beam or part of the design, where they are introduced to facilitate the SMS headlamp designs. process. The effect of these extra surfaces is taken into account by Keywords: SMS Design, Freeform Optics, Illumination, LED collimator, simply propagating the source and target wavefronts through those RXI, Nonimaging Optics, TIR-R, Diamond Turning, SMS Imaging, surfaces. The resulting wavefronts, now refracted or reflected at the Condenser non SMS surfaces, are then used in the SMS calculation. Besides the obvious physical limitations (i.e. etendue conservation [1]), the only Introduction condition for a successful SMS design process is the absence of any The SMS design method gives access to solutions of optical problems caustics of the wavefronts to be coupled in the vicinity of the space to that can be formulated as the coupling of incoming and outgoing be occupied by the SMS surfaces. wavefronts. The number of wavefronts that can be controlled on both In the following example, the edge rays of a 2D light source are to be the input and output side, depends on the number of surfaces to be coupled to two straight exit wavefronts. If the angle between those exit designed. Other methods create only one profile or surface at a time wavefronts is chosen according to the Etendue of the light source and and therefore only couple one wavefront to another. This can be the exit aperture size, this system behaves like a perfect 2D collimator. understood as a generalized Cartesian oval construction. Those solutions The input parameters are: Two sources and two target wavefronts, a may be perfect for point light sources but, in many practical cases, the starting point for one of the SMS surfaces and its normal, the nature of approximation of an extended light source as a point source results in the surface (here refractive) with the index of refraction and two optical loss of efficiency and “smearing out” of the output pattern. Iterative path lengths between the conjugated pairs of wavefronts. design methods try to optimize point source solutions for extend sources. In contrast to this, and without the need of optimization loops, the SMS method inherently uses the extension of the source as an input parameter to create an optical system perfectly adapted to it. Such systems exhibit performances close the thermodynamic limits. While the SMS 2D method has generated a number of ultra compact and highly efficient devices with rotational or linear symmetry, SMS 3D has matured into a stable design method for a wide range of applications Figure 1: Schematic of design procedure for an RR SMS 2D collimator. The two profiles are calculated starting at a first point P0 (left). All other points follow from this start point in an alternating algorithm.

The profile of the gray lens in Figure 1 shall be designed. The first point P0 and its normal n0 are chosen as input parameters. A ray from the first source wavefront (WFi1) is selected that connects the edge of the source and P0. This ray is refracted. The first exit surface point P1 is found by searching a ray that, emitted from the exit wavefront WFe1, has an optical path length L1. Now a ray from the other exit wavefront WFe1 is refracted at this point, and traced. Point P1 is calculated the same way as before, now utilizing the optical path length L2. Repeating the same steps produces alternating points on the input and exit surface. On the far right of Figure 1 it can be seen, that all rays from each edge of the source leave Figure 2: Schematic design procedure for an RXI SMS 3D collimator. Left: “Vertical” wavefronts of source and target are coupled to create two vertical profiles. One profile is the seed curve used to calculate the SMS chains the lens as rays normal to the exit wavefronts. From the edge ray theorem (right). it is known, that all other emitted rays will not be edge rays and therefore have a smaller off axis angle as the design rays. In total three input wavefronts (e.g. three corners of the chip) and three exit wavefronts are chosen that describe how the light from the chip The final lens profile is obtained as the minimum order interpolation corners shall be emitted. In the example in Figure 2 all exit wavefronts curve through the SMS points. While strictly speaking the method only are planes rotated 5 deg from the z axis, WFe1 and WFe2 horizontally controls rays from the two design points, the behavior of rays emitted and WFe3 vertically. This RXI shall produce a far field pattern that is an from other points than the edge points as well as all rays hitting the image of the chip of 5 deg vertical and horizontal size. In a first step (Fig lens profile where the profile is interpolated between the SMS points 2, left) two vertical SMS profiles are calculated that couple the two are not controlled. In most practical cases, however, SMS designs are “vertical” pairs of wavefronts with each other. One of those profiles will “well behaved” in the sense that, firstly, the interpolated profile sections be used as a seed curve for the calculation of the SMS chains (Figure 2, are not distinguishable from a “perfect” profile for rays originating from right): Each pair of chains is calculated by coupling the two pairs of the design points, and secondly, that all rays emitted from source points “horizontal” wavefronts with each other. The seed curve can be sampled that are not design points behave as expected. at as many points as needed in order to calculate as many chains as desired. The two families of chains are then approximated (Figure 3, SMS 3D Review left) by a smooth CAD surface. The RXI shown now directs all rays from The SMS 3D design method has, as the 2D design, as input parameters the two lower corners of the chip to 5 deg Left/ Right and, in this two pairs of wavefronts, now defined as 3D surfaces and two optical path example, to 0 deg horizontal. Rays from the third design point (an upper lengths between the two corresponding pairs of wavefronts. But, instead corner of the chip) that hit the SMS surfaces near the seed curve will be of a single starting point and a normal as free parameters, now a “seed” directed 5 deg down while all other edge rays from the LED chip top curve in space can be chosen that will serve as starting points for the corner theoretically are not controlled. However, in many applications, SMS point calculations. SMS points generated using the 4 design SMS designs are sufficiently well behaved to maintain the desired wavefronts from a point on the seed curve are called chains. The seed characteristics for all non design points (Figure 3, right). curve can be sampled at as many points as desired to create many chains. The full design is eventually defined by all the SMS points that can be The design shown collects the source light within a half-hemisphere. interpolated by two 3D surfaces; one of them contains the seed curve. The other half of the light (the light the chip emits upwards) would be captured by a second design procedure done similarly. The two individual In the definition of the seed curves lies an important degree of freedom: RXIs can collect 100% of the light emitted by a light source into a It can be obtained by a SMS calculation using for example WFi1/WFe1 hemisphere or even up to higher off axis angles. and two new wavefronts WFi3/WFe3. This third wavefront pair can consist of a source wavefront emitted from a third point of the source coupled to a third exit wavefront. The full 3D SMS design maintains the coupling of the third wave front pair in surface regions in the vicinity of the seed curve. In surface regions that are not close to the seed curve, this coupling may be lost. This means that a 3D SMS design can, with two calculated surfaces, control three design points, two of them perfectly, one of them approximately. Imagine a square light source (e.g. an LED chip, Figure 2). The SMS device shall be a dielectric solid, with a metallized back surface (the “X” surface) and a refractive front surface, that will at the same time reflect Figure 3: RXI SMS 3D collimator. On the left the two 3D surfaces of the RXI and some source images are shown. On the right: Raytrace of a “well behaved” SMS design with three planar exit wavefronts. rays by TIR (called “I”) and refract others (called “R”). Those designs are called RXI [9] [10].

Comparison of Conventional SMS Conventional design methods design one surface at the time. An example is a freeform mirror used for automotive head lamps. The curvature of every mirror point determines the size and position of the image projected from a small surface element, often called a pin-hole. One design method consists of obtaining a single free-form refractive or reflective surface [4] [5] [6] to solve a prescribed-irradiance problem based on the small-source approximation. This strategy (usually called Figure 5: Left: Photo of TIR lens produced by LPI. Right: Working principle of TIR-R combination. point-to-point mapping [7]) is well known either for rotational optics, where its solution simply involves the integration of a non-linear The real-world efficiency of the optical system injection molded with ordinary differential equation [8] or for the more general 3D design of PMMA strongly depends on the surface quality (roughness and a single free-form surface, requiring the solution of a non-linear partial deviations from theoretical profile) and the teeth radii of the TIR lens. differential equation of the Monge-Ampere type. All of those solutions Since both the convex and the concave vertices are hit by the incident work well in the small source limit case. sunlight in the solar application, vertex rounding will reduce the usable SMS designs control two source points perfectly and a third one surface area and therefore the power transfer to the solar cell. partially. The SMS design is fully determined by the choice of the wavefronts and some other simple input values, as described above, with the need of optimization steps.

Figure 4: Far field images created by a “pinhole” (upper two graphs) or a set of pinholes (lower two graphs) on the exit surface of a free form optical system illuminated with an extended source. Left: Conventional design: No Figure 6: Left: Photo of a TIR lens section and details form its convex and concave tooth vertices. rotation and distortion control of the images, Right: SMS Design: Size and orientation control of images. The TIR shown in Figure 5 and 6, designed for a solar concentrator, has A conventional design only controls one point of the source images. small vertex angles (around 28º) and a diameter of about 40mm. LPI, Different sections of e.g. a freeform reflector create far field images of who owns the IP for and manufacturers the TIR lens, has demonstrated the source of varying size, orientation and distortion (Figure 4, left). The extremely small tooth radii for lenses molded in a prototype tool. The rotation of the images and their distortion may, if the images are large molding tool has been produced by diamond turning. Injected lenses or if the features to be obtained in the pattern are small, seriously have been cut to examine their profile under a microscope with digital degrade the quality of the output pattern. Not so in an SMS design: The x-y readout. The convex tooth radii have been measured as 5μm in pinhole images of the source are guided by 3 source points in a average but some radii are below 3 μm, so that they cannot be faithfully deterministic manner, so that orientation and size can be controlled. resolved under the microscope. The concave teeth yield 14μm on average and 5μm minimum. Such small radii suppress the efficiency Examples of SMS Applications losses of the system due to rounded teeth edges below 1%.

TIR-R solar concentrator and LED collimator RXI LED collimator The TIR-R system [11] consists of a TIR lens with one flat surface and a The development of the rotational symmetric RXI is a very good example secondary lens, both rotationally symmetrical. The secondary lens has a of the SMS 2D design method. Its striking characteristics have been cavity that, filled with an optical coupling gel, holds either a receiver described elsewhere [10]. In Figure 7 an application as an MR bulb (e.g. a solar cell) or an LED chip as emitter. The two SMS surfaces replacement is shown: The RXI collects and collimates the light of designed are the reflecting surface of the TIR rings and the surface of LumiLeds’ Luxeon Lambertian LEDs of any color. The lens is fitted into a the secondary lens. The system has been developed for solar applications housing that includes drive electronics (input voltage 6-24Vpeak AC/ to concentrate the solar light onto a small high efficiency solar DC) and the LED to make it a full bulb replacement. The RXI lens achieves photovoltaic cell. The theoretical efficiency is 82%. The same optical an extremely tight beam with an opening angle at the physical limit set system, with an adjusted design, can be utilized to collimate LED emitted by the Etendue of the source. The lens diameter is 36mm and its depth light with 85% efficiency. is only 9mm.

By varying the parameter set that our program uses to calculate the RXI surfaces and by choosing outgoing wavefronts, that are adapted to the low and high beam pattern, we were able to obtain “legal” designs for both the low beam and high beam RXIs. In detail raytraces, the designed elements prove their performance. The low beam yields total optical efficiencies (defined as the ratio of flux projected onto the road in +/-30deg horizontal and +/-10deg vertical window divided by the full source flux) of 80%, assuming a reflectivity of 93% for the metallized surface, and anti reflection coatings on the refractive surfaces. Without Figure 7: Left: Photo of a “Chip LED bulb replacement”2. An identical RXI as mounted in the housing is held in front of the full assembly. Right: Radiation pattern of the RXI with a green Luxeon LED. The vertical bars visible anti reflex coatings 75% (68% with cover lens) is reached. in the pattern are caused by the contacts of the LED chip.

The RXI is designed as a nonimaging device to ensure highly efficient light transfer. This is not incompatible with a good image formation although the names imaging and nonimaging suggest an antagonism [15]. The radiation pattern of the RXI in the far-field therefore shows a clear image of the LED chip.

RXI LED headlamp designs The RXI’s design principles laid out in the SMS3D review chapter in this paper, have been applied to design several LED headlamps. These are the first Free-form RXI devices ever designed. Such a design will only be Figure 8: Schematic view of a 3-D RXI (for low beam). efficient, if it collects light from the entire solid angle of the source emission, collimates the light and forms the beam pattern only by The high beam RXI is even more efficient: Up to 85% with and 80% reflection or refraction without the help of any blocking devices or without anti reflex coating can be achieved. The low beam RXI meets shutters. Very high efficiency is important because high performance the regulations FMVSS108 Table 17 with 8 LEDs (and 8 RXIs) of 80lm LEDs are expensive light sources compared to standard light bulbs and each. It also passes the legal gradient values as defined in FMVSS108 the available LED flux still is at least one order of magnitude lower than easily. For the high beam, an additional 8 LEDs of 80lm are used. Both that of an ordinary bulb. The RXI is highly efficient, as well as compact sets of RXIs together meet the high beam specifications. The hotspot of and optically pleasing, an important “sales tool” in the automotive field both functions combined is 50000 cd. Prototypes for both low and high where styling is paramount. beam have been manufactured that perform with almost the theoretical Elaborate calculations carried out by proprietary software create the efficiency. Values of 75% efficiency for a diamond turned low beam two free from surfaces that transform the source light into the desired element without anti reflex coatings and 84% efficiency for an injection pattern without the use of faceting, shutters or any additional optically molded high beam element with anti reflex coatings have been active surfaces. Extensive raytracing with in-detail modeling of the measured. Raytraces had predicted efficiencies within 2% of the light source and prescribed optics verify the designs. measured values. The device shown in the following section represents a typical design of The low- and high beam units can be arranged in many different an RXI 3D collimator. RXIs for the low beam and high beam application configurations to meet designer’s expectations. The final number of RXI both for the US market (SAE) and the European market (ECE) have been modules is inversely proportional to the available flux of the LED. designed using the 3D SMS method. The dimensions of the device are By choosing different design parameters and wavefronts, an RXI similar dictated by the etendue calculations (to limit the vertical or horizontal to the shown design in appearance and size, has been designed to meet extent of the pattern) and luminance conservation to achieve the ECE low beam and high beam specifications. Because of the more needed hotspot intensities. The RXI has a non active center region demanding gradient specifications of the European regulations and the because the two reflections of the rays displaces them several lower stray light levels (light projected above the horizon), those designs millimeters up- or downwards from the center of the light source. This must exhibit even better light control than the SAE versions. The SMS will increase the device size, as the lit aperture size is smaller than the design has therefore been refined to create more precise 3D surfaces physical aperture of the RXI. A possible design that has been investigated that reduce ray angle deviations from the theoretical wavefronts used 3 is approximately 27 x 60 x 15 (w x h x d) mm in size for the low beam, for the design. Raytraces demonstrate the success of those designs; 3 where the high beam RXI (not shown) is about 40 x 55 x 17 mm . The however, as the gradient is formed by imaging the LED chip edge, high beam RXI is wider to meet the hot spot requirements and to better placing tolerances on the luminance distribution of the LED chip within collimate the light in the horizontal direction. its package, the RXI lens must be very tightly controlled.

Outlook onto future LED lamps designs most compact collimator, but it also uses metallic reflection. For mass A single LED Headlamp will be possible as soon as there will be a chip production, besides the fact that no more then 90% reflectivity at each available that produces more than 500 lm. In the hypothetical case that ray interaction is achievable, this metallization means added cost. an LED with chip dimensions similar to today’s LumiLeds Luxeon LED TIR reflection is the most efficient reflection process and it doesn’t would be available, a full low beam headlamp could be of the size of a require additional coatings. Today’s modern tool making (diamond single RXI of 25 x 55 x 14mm3. Smaller LEDs with the same flux would turning) and injection molding processes enable the reproduction of lead to a linear reduction of aperture size. However from a safety and optical plastic surfaces that achieve the necessary low surface design standpoint too small aperture sizes may not be desirable. roughness and flatness to ensure full TIR. Several LED collimators using Besides the inherent advantages of LEDs (their longevity, instant turn TIR exist on the market, but they often exhibit low efficiency and high on, “whiter” color, and smallness), very bright and small LED chips depth, so that the associated large plastic volume makes injection slow enables the optical designers to create beam patterns of unsurpassed and expensive. Also, their minimum diameter is relatively large if high beam quality, in terms of brightness, uniformity and light control. The collection efficiency is to be maintained. SMS method offers the necessary precise light control needed to realize Here we present a new class of LED collimators, called “UFO”, that those future designs. drastically reduces the plastic volume. The small wall thickness speeds up plastic injection cycle times and lowers cost. At the same time the Incandescent SMS headlamp design is extremely efficient: Raytracing results have demonstrated Another successful application of the SMS 3D method it the design of efficiencies as high as 92%, so that all losses can be attributed to a low beam and high beam headlamp that, as conventional headlamp Fresnel reflections at the input and exit surfaces of the device. If anti designs, uses a halogen bulb as a light source. The optics, however, reflection coatings were applied, efficiencies of almost 100% could be doesn’t consist of a typical single free form reflector or a projector achieved. system as today’s headlamps, but of a complex lens, that redirects the light from the source by refraction as well as by metallic mirror surface and total internal reflection. As the optical system envelopes the light bulb, the collection efficiency is much higher than of conventional head lamps. Because of the perfect image control of the SMS method, all filament images are kept horizontal all over the exit surface of the optical system. This allows us to reduce the vertical height of the system in comparison to conventional designs. The result is a headlamp that is more compact both in exit aperture size and depth and more efficient Figure 9: Schematic views of a UFO LED collimator, on the left with a Luxeon emitter and ray trajectories, on the right with a plastic holder to place the LED over the standard Luxeon Star LED. in terms on flux projected on to the road than today’s best halogen bulb headlamps. The UFO design utilizes several sections of different ray trajectories. All Because of the proximity of some parts of the design to the hot light surfaces have tailored profiles. The central section captures the LED bulb, glass has been chosen as lens material. In consideration to our light emitted close the optical axis and collimates it by two refractions. customer, we are unable to reveal more detail on this development at The second section, of greater off axis angles, is an “RIIR” section: The this time. light is first refracted, then reflected (TIR) at the exit surface, another time reflected (TIR) at the back surface and then refracted at the outer UFO LED collimators section of the exit surface. Both reflections are TIR. The third section High power LEDs (e.g. the LumiLeds Luxeon LED) today put out up to works as “RIR” (Refraction, TIR, and Refraction) and collects the light 80lm of white light. Those flux values are made possible by relatively emitted to large angles. large LED chips that, in the case of the mentioned LED, measures 1 x 1 A variety of designs have been derived with diameters as large as 36mm x 0.1 mm3. Because of the Etendue conservation, it is not possible to to achieve very narrow collimation and as small as 14mm, much smaller incorporate a collimator into an LED dome (e.g. of mm diameter) that than it would be possible for conventional collimators without would at the same time be efficient and produce a small collimation compromising efficiency. As a standard product, to be marketed soon angle of the output beam. Therefore, a round dome is placed over the by LPI, a diameter of 17mm has been chosen, as a compromise between LED chip and the collimation of the now near Lambertian light source is collimation angle and compactness. (For comparison: the Luxeon Star left to external lenses. Purely refractive systems are unable to cover the LEDs are provided attached to a hexagonal PCB of 18mm width). The full solid angle of the emitted light. The collimation of the LED is a lens itself is only 6.6mm tall, making very compact assemblies possible. classical problem of nonimaging optics- for example the CPC [1] is a Comparable LED collimators available on the market have slightly larger well known solution to this problem. However, mass producible metallic optical apertures (approximately 18mm). They are approximately 10mm surfaces don’t reach very high efficiencies and CPCs, for small tall (see Figure 10), almost 50% taller than the UFO. Their plastic volume collimation angles, are very deep. The aforementioned RXI is today‘s is almost three times higher than the UFO.

illuminated differently in the two positions of the lens by the three ray bundles of the UFO. The efficiency of the device varies between 80% and 90% at wide and narrow beam positions, respectively. Even between the two different design positions, the device doesn’t show excessive intensity variations in the form of concentric artifacts. The central part Figure 10: Schematic views of two commonly available Luxeon collimators (left and center) and the LPI UFO (right). All three designs have similar exit apertures and collimation angles. The two designs on the left have of the distribution shows a continuous drop in intensify towards the plastic volumes of approximately 1500 and 1300 mm3 respectively, while the LPI UFO (right) has a volume of wide beam position without any dark center formation. about 500 mm3.

While the collimation angle of the UFO for the different Luxeon LED colors varies between 8° and 10º FWHM, the efficiency is 90% for all Lambertian Luxeon LEDs. In order to place the UFO lens over the LED and onto the PCB, various holders have been designed that fit over the Luxeon star 1W and 3W, and can also fit directly onto a PCB if the naked Luxeon emitter is used. Different UFO models for wider collimation angles but with the same Figure 12 Left: Photo of UFO Zoom lens. Center: Raytraced simulation of the narrow beam with 6º FWHM. Right: Wide beam simulated with flat intensity and 80º FWHM. Raytraces are represented in Munsell “lightness” coding. footprint, size and exit surface are being developed by LPI. Those designs exhibit flat intensity patterns and well defined cutoff angles. Future SMS Applications

Compact SMS 3D integrators Integrators have been employed to homogenize the illuminance of projection systems. Such integrators normally consist of two arrays of micro lenses that have the same focal length and are at the same time separated by their focal length. The intensity distribution of the second lens is an image of the illuminance distribution on its focal plane. If the lens pitch is small compared to the distance from the source, the illuminance to be imaged by one micro lens will be almost constant, so that the outgoing intensity pattern will also be uniform. The integration Figure 11: Photo of other UFO LED collimators, designed for specific applications. The lens on the left has a diameter of 18 mm, the right one 30 mm lateral dimension. zone can be designed to include the entire emitting zone of the source and the output will be independent of any luminance variations in this zone. If the integrating zone is chosen larger than the source size, the Single piece zoom optics UFO LED collimator output will stay unchanged, even it the source moves within this zone, For various illumination applications zoom lenses are employed. so that the design can be made tolerant to source positioning errors. Classically this is solved by moving the incandescent bulb along the optical axis out of the focus of a parabolic reflector. However, in most As SMS designs work close to the light source, they may be more cases neither the focused position forms a bright, well defined and dependent on positional tolerances and the illuminance distribution of homogenous hotspot, nor is the wide beam free of radial intensity the source than conventional designs that use larger optics farther variations noticeable as dark or bright concentrical artifacts. Moreover, away from the source. In order to combine the compactness of an SMS if a source is moved out of the focus of a parabola, in many cases a dark design with the robustness of a conventional design, we have designed center spot occurs. free form micro lenses that can be combined with SMS designs without the need of extra integrating elements. These designs can integrate in In order to create a LED collimator with a zoom function, a secondary one or two spatial directions. A typical application where the integration lens, after or before the collimator, can be employed, that can either in one direction is desirable is the automotive headlamp, where a well rotate or move laterally or longitudinally to open the narrow beam to defined vertical cut-off has to be achieved. In the case of the ECE form a homogenous wide beam. We have employed a different concept: specification, the intensity may have to drop from 30,000cd to almost We use the UFO itself to perform this function without the need of 0 in little more than 1deg vertical variation, but the horizontal pattern extra elements. is very wide and has a smooth roll off. An SMS integrator design can A particular design example has a diameter of 35mm. The UFO is produce a vertical cutoff that is independent of the source illuminance designed in a way to work as a collimator in one position, and, when distribution and, depending on the design parameters, also tolerant to moved 2mm away from the Luxeon LED, opens up the beam in a very vertical source positioning. homogenous way with almost perfectly flat intensity in the far field up to 40 º off axis angle. The design uses tailored surface sections that are

A different application, where integration in two spatial directions is SMS imaging necessary, is the color mixing of light from an RGB light source The SMS method has been developed for nonimaging applications but consisting of three or more sources that are distributed over different the method can be applied to imaging solutions: If the input and exit positions (e.g. an RGB LED with 3 chips in one housing). We expect SMS wavefronts are spherical (for object and image points in the near field) integrator devices to be developed in the near future that collimate or flat, for points at infinity, one pair of points will be perfectly imaged light from such an RGB source while maintaining the same intensity into two image points. The well known Cartesian converts a parallel ratio amongst the three sources and therefore the color of the combined wavefront into a spherical wavefront. A single surface that converts light at any point in the intensity pattern. one general input wavefront into a given exit wavefront can be understood as a generalized Cartesian oval [14]. It is not trivial, however, SMS condensers that two surfaces (e.g. of a lens, see Figure 14, center) can convert two In digital projection the most commonly used light source is an arc inputs into two exit wavefronts, and that n SMS surfaces will image n discharge lamp. The full optical train consists typically of light source, points into their conjugate points. With many points imaged perfectly, elliptical mirror, glass mixing rod to achieve constant illuminance at the one can hope that all adjacent points, and eventually a complete area exit surface of the prism, LCD micro display and optics to project an inscribed into the design points, may be imaged with very good quality image of the LCD display onto a screen. The achievable brightness onto the imaging plane. Conventional optics, consisting of spherical depends on all elements but the condenser may be the most critical. surfaces, doesn’t image even a single point perfectly because of spherical aberrations. In order to improve a lens design, often some The typical elliptical mirror collects a certain fraction of the fully emitted paraxial aspheres are added. Conventional optics, however, fail to work light and forms arc images on the entry aperture of the light mixing rod. well for non paraxial rays because the optimization methods start off The size of the images varies (in meridian length and sagittal width) from the paraxial regime and spherical surfaces. In some cases large off from point to point on the condenser exit surface and the images rotate axis angles are desired, e.g. the imaging section of projectors in order to due to the rotational symmetry of the mirror and the arc. Due to the minimize the projector-to-screen distance. Conventional lenses are rotation of the arc images, conventional condensers perform worse limited to projection angles of about 30-40º. when the target aperture is not circular but rectangular, especially when the aspect ratio is high as for a 16:9 picture format. A mixing rod of larger diameter would increase the efficiency but this would have to be combined with a larger display, at a prohibitive cost.

Figure 14: Left: A Cartesian oval focuses a parallel ray bundle into a point. Center: An SMS lens can focus two wavefronts into two different points. Right: Two SMS lenses can focus four objects into four imaged points.

SMS 2D concentrators have been found to work as ultra-high numerical aperture imaging devices [15]. Also, achromatic aplanatic aspheric

Figure 13 Digital projector condenser, left: Conventional design with arc imagines of varying size, on the right an doublets [16], designed with SMS 2D, have been demonstrated. In improved design with constant arc image sized using a tailored reflector and glass cover profile. general, SMS lenses don’t have any paraxial limitation and can therefore be designed to large off-axis ray angles. A projection system with SMS As en ellipse perfectly transforms rays emitted form on of its focal surfaces could therefore be extremely compact as projection angles up points onto the second focal point, all other points are not controlled. to 85 deg can be realized. An SMS 2D design controls two points (the extreme ends of the discharge in this case) that can be mapped onto a circle of the mixing The possibilities of SMS 3D in imaging optical systems are still being rod entrance aperture to control the meridian image size. This design evaluated but the added wavefront control can make these much more (Figure 13) would use the mirror surface and one of the surfaces of the powerful than SMS 2D designs. However, the resulting free form lamp glass cover as SMS surfaces of rotational symmetry with tailored surfaces may be a challenge if applied to glass optical elements but in profiles. A similar solution has been described before, although in the reflector designs, production technology has reached a level of accuracy limit of a small source [1] [13]. Such a design results in a noticeable that makes free form imaging possible. increase of in-coupling efficiency. However, from the description of the SMS method above, it should be clear that a full SMS 3D design would be even more beneficial as it also controls the rotation and the sagittal image extension. Such a design would not have rotational symmetry but it can project source images of constant positional and aspect ratios exactly tailored to the mixing rod dimensions.

Summary The SMS 2D design method has been applied for many practical devices, some of which are presently being mass produced. The much more complex SMS 3D method has reached a maturity and unsurpassed light control capability that makes it the tool of choice for complex illumination tasks and, in the near future, for certain imaging problems. The SMS 3D method, applied in several automotive LED headlamps designs, has proven to produce highly detailed intensity patterns, while at the same time reaching the highest efficiencies and ultra compact dimensions. A future stage in SMS 3D will be the design of three freeform SMS surfaces to perfectly control three wavefronts to enable even a higher level of control of the output beam quality. SMS 3D devices are expected to be mass produced soon and will appear in many different applications ranging from illumination to collimation, condensers and image projection.

References : All optical devices shown are protected by patents awarded, in allowance or pending. See under patent list. 1. W.T. Welford, R. Winston. “High Collection Nonimaging Optics”, Academic Press, New York, 1989 2. P. Benítez, R. Mohedano, J.C. Miñano. “Design in 3D geometry with the Simultaneous Multiple Surface design method of Nonimaging Optics”, in Nonimaging Optics: Maximum Efficiency Light Transfer V, Roland Winston, Editor, SPIE, Denver (1999). 3. P. Benítez, J.C. Miñano, et al, “Simultaneous multiple surface optical design method in three dimensions”, Opt. Eng, 43(7) 1489-1502, (2004) 4. L.A. Cafarelli, V.I. Oliker, “Weak solutions of one inverse problem in geometrical optics”, Preprint, (1994) referenced in L.A. Cafarelli, S.A. Kochengin, V.I. Oliker “On the numerical solution of the problem of reflector design with given far-field scattering data”, Contemporary Mathematics, Vol.226(1999), 13-32 5. S.A. Kochengin, V.I. Oliker, O. von Tempski, “On the design of reflectors with prespecified distribution of virtual sources and intensities”, Inverse problems 14, pp.661-678, (1998) 6. H. Ries, J.A. Muschaweck, “Tailoring freeform lenses for illuminations”, in Novel Optical Systems Design and Optimization IV, Proc. SPIE 4442, pp. 43-50, (2001) 7. W. Cassarly, “Nonimaging Optics: Concentration and Illumination”, in the Handbook of Optics, 2nd ed., pp2.23-2.42, (McGraw-Hill, New York, 2001) 8. W.B. Elmer, “The Optical Design of Reflectors”, 2nd ed. Chap. 4.4 (Wiley, New York, 1980) 9. J.C. Miñano, J.C. González, P. Benítez, “RXI: A high-gain, compact, nonimaging concentrator”. Applied Optics, 34, 34 (1995), pp. 7850-7856. 10. R Mohedano, J. Miñano, P. Benitez et al, “Ultracompact nonimaging devices for optical wireless communications”, Opt. Eng. 39 (10), 2000 11. J. L. Álvarez, J. Miñano, P. Benitez et al, “TIR-R concentrator: A new compact high-gain SMS design”, in Nonimaging Optics: Maximum efficiency light transfer VI, Roland Winston, Editor, Proceedings of SPIE Vol. 4446 pp. 32-42, (2001) 12. JP71749740, assigned to Seiko Epson 13. US Patent 5,966,250, assigned to Philips 14. R.K. Luneburg, “Mathematical theory of Optics”, (U. California, Berkeley, 1964) 15. J.C. Miñano, P. Benítez, “Ultrahigh-numerical-aperture imaging concentrador”, J. Opt. Soc. Am. A/ Vol. 14, No.8 (1997) 16. J.C. Miñano, P. Benítez , F. Muñoz. “Application of the 2D etendue conservation to the design of achromatic aplanatic aspheric doublets”. Nonimaging optics: maximum efficiency light transfer VI, R. Winston Ed., Proc. of SPIE Vol. 4446 pp. 11-19, 2001

Patent List: SMS 2D Design method, RXI 2D (rotational symmetry): 2D SMS patent (issued): “High Efficiency Non-Imaging Optics” Patent # 6,639,733, Issued Oct. 28, 2003. Inventors: Juan C. Minano, Pablo Benitez, Juan C. Gonzalez, Waqidi Falicoff and H. John Caulfield. RXI 2D (rotational symmetry), UFO, UFO Zoom: Air-gap RXI and CIP continuance (in allowance): “Compact Folded-Optics Illumination Lens”. Inventors: F. Muñoz, Juan C. Minano and Pablo Benitez. SMS 3D Design method, RXI 3D, Incandesent SMS 3D; 3D SMS: “Three-Dimensional Simultaneous Multiple-Surface Method and Free-Form Illumination-Optics Designed Therefrom” (pending). Inventors: Pablo Benitez and Juan C. Minano. TIR-R: J.C. Miñano, P. Benítez, “Device for concentrating or collimating radiant energy”, PCT/ES00/00459

Plastic Optics Enable LED Lighting Revolution > Tomi Kuntze, Managing Director, LEDIL Oy

Only some years ago LEDs were predominantly used as tiny marker lamps in electronic devices. The recent rapid development in semiconductor and packaging technologies has quickly turned LEDs into powerful lighting components, capable of replacing existing light sources, such as bulbs, halogen lamps or even highly efficient fluorescent tubes. So far, the adoption of LED technology has been strongest in architectural lighting, signage, automotive lamps, mobile phones and displays. In everyday life, more and more public places such as building facades or bridges are being illuminated by colorchanging LED systems. The next major area for penetration of the LED technology will be in general lighting. An essential enabler for LED lighting revolution over the coming years is a clever use of optics to collect and shape the light from the LED Figure 1: LEDIL CAT lens for Street lighting. package. In the following article we go into the details of how to do it by means of advanced plastic optics.

Illumination Engineering Applications of optical engineering related specifically to illumination systems are often called illumination engineering. As illumination engineers, who wish to design a good LED lighting system, we need to pay attention to many of the following important issues:

• Which LED and which color is most suitable for this application? Figure 2: LEDs as light sources vary much from each other. The same optics cannot be used for several LEDs without jeopardizing the performance. • What is the optimal type of light distribution for this application? • What is the CRI value needed – is brightness more important than a good CRI (Color Rendering Index, describes color appearance)? • How uniform shall the illumination be, both in terms of brightness and color, over the object to be illuminated? Do you allow any shadows or other inconsistency in the projected beam of light? How consistent shall the quality of the light (intensity, color) be over time? • Do we need to choose any special materials or components due to the environment where the system is assembled in? • Do we need any modularity of the system components? • What kind of flexibility is needed from the system setup, to enable assembly of it next time in another place? Figure 3: LEDIL FLARE lens for Luxeon K2 and LEDIL FLARE-C lens for Cree XR-E show how different light sources lead to different lens design, with the same illumination specification. The list above can be endless, which shows the complexity of requirements for good illumination engineering. It also means that a LED lighting system and its components always are results of various Challenges with LED as a Light Source compromises. There seldom is a way to fulfill all the requirements at the Even as LED packages have developed much during recent years, there same time. If it was done, it would often mean a system that is too does not seem to exist any common idea among LED manufacturers, complex to use, too difficult to manufacture or too expensive. Therefore, what an optically good LED package should look like. All of them are let us take a look at the various aspects of a best possible compromise, different from each other and only a few of them have been designed for each in turn, and try to figure out, how we can create a best possible easy adoption of secondary optics. Many of them have been designed to optical system for LEDs. get a brightest possible beam out of the LED itself, without thinking how the beam can be used in the following step: secondary optics.

In all LED packages the chip has a square or rectangular shape. If a conventional, highly efficient lens is used and this lens has been designed with an idea to make a most accurate image of the object in its focal point, the illumination result is the same square or rectangular shape of the LED, with all chip details projected on the object to be illuminated. Nobody likes it, as the first rule for an illumination engineer is to make a beam that is smooth and pleasant. Therefore, the challenge is how to maintain the high efficiency of a conventional lens, but at the same time get rid of the square and make a round shape instead. Another challenge is, how to achieve an even white color on the object, even if the LED used does not emit uniformly white light? The problem Figure 4: Light distribution of a lens, with the light source it has been optimized for (left), and the same lens, here may be a result of variations in the blue LED underneath the positioned at a correct focal point, but using a LED, which the lens has not been designed for (right). phosphor layer, unevenness of the phosphor layer or some die manufacturing details. In many cases, a secondary optics has to heal problems that actually are LED related, but become apparent when the LED is exposed to a secondary optical system. A third main challenge for optics is multiple-die LEDs, which have a great number of chips, populated densely, close to each other and covered with a common phosphor layer or dome. It is very difficult to develop optically efficient and visually pleasing optics for such LEDs. The reason is that most illumination optics development is based on an idea of a point light source. Having e. g. 4 x 4 dies makes it impossible to use this principle in design, as dies are too near to each other to be seen as separate, but too far away from each other, to be seen as a point light source. The final result is that such light LEDs cannot be used in applications where a good collimation and narrow light distributions Figure 5: An example on the use of a clip-on sub lens to change illumination characteristics of a lens. are needed. Main benefits for LEDs, when compared to other light sources, are the directional light they give and the relatively small size of the light source. In that sense LEDs easily outperform all other light sources, such as bulbs and fluorescent tubes.

Optimization with Regard to Light Source Characteristics

Some companies have gone so far in the standardization of their lenses Figure 6: An example of modular lens system, with adhesive tape fastening, is the square lens series with more that they have decided to treat all LEDs as similar light sources and than 50 different versions (light distributions and optimized for various LEDs). offer the same lens for many LEDs. In my opinion, this is not possible, nor rational to do, as every LED is very different from each other. As a Basic Optic Types clear evidence of this difference I show 3 different simulation pictures, LED optics can be categorized in e. g. lenses, reflectors, combinations of which show the illumination result, when a lens, optimized for LED ”A“ them and systems of multiple lenses or reflectors. It cannot be claimed is put at its correct focal length, but used in combination with lenses that some of them would be better than another. It fully depends on the ”B“ and ”C“. You can see how the well controlled beam in ”A“ changes to application and its requirements. a scattered plot in ”B“ and ”C“. Generally speaking, however, lenses are more efficient in shaping the E.g. LEDIL designs and optimizes every optical system with regard to the beam than reflectors. The main reason is that light in a lens travels LED it is going to be used with. It guarantees that a full efficiency both through at least 2 surfaces (often more), before getting out of the in terms of power and visual appearance is achieved. system in the desired angle, while for simple reflectors a part of the light gets out without ever hitting any optical surface. On the other hand, every time light enters or exits a surface it looses its efficiency, which speaks against complex lens systems and favors reflectors.

Most companies offering LED optics base their optical idea on collimating Another important feature is the shape of the components and the the light from the LED with a standard collimator lens design, spreading easiness of integrating them in the surrounding mechanical systems of the light on the top surface of the lens in the angle needed. The quality an illuminator. As an example of a standard platform and shape is the of the resulting beam depends on how accurately the LED is studied, LEDIL square lens series, with 50 different optical units with the same and the degree and method of inter-connecting the optical surfaces to size and shape. Furthermore, this component can also be supplied as each other. separate parts, which enables the customer to replace a part of e. g. the lens holder construction by the structure of his illuminator. Free Form Lenses In general terms, all kinds of mechanical features can easily be integrated A more advanced idea is to freely modify all the optical surfaces, when in a plastic optic – let it be a lens or reflector or a group of them. As having an accurate model of the LED and knowing what the resulting examples of features used in several lenses are: adhesive tapes, threads, beam shall look like. This kind of optical system is called free form snap hooks, pins, screw holes and welding features. system. Free form does not mean that a computer calculates the optical geometry with given input and output conditions. For good free form Optical Materials optics there is also a structured, wellconnected underlying optical Without going deep into the materials science, it is essential to give a concept, which the illumination engineer in charge has created. Clear quick note on materials to be used in advanced optical systems. There benefits with free form optics are e. g. good modeling of LED resulting is a difference in materials, which often cannot be seen with bare eyes. in great accuracy, high optical efficiency, controlled beam quality and a Most of the LED lenses of today are manufactured of PMMA (acrylics) smaller physical size. or PC (polycarbonate). There are special applications requiring COP or PA12 materials, but they are rare. A special attention should be paid to Reflectors the variation within materials supposed to be ”PMMA“ or ”PC“ only. Reflectors have been used as optical element for a longer time in history There are hundreds of different materials in these material groups and than lenses. Reflectors are easy to manufacture, therefore cheap. A only a low percentage of them fulfill the criteria of being excellent for common denominator for all reflectors is a relatively large opening advanced plastic optics. Care must be taken of to insure good angle, i. e. a large part of the emitted light from the LED gets out of the dimensional accuracy and stability, heat resistance, UV resistance, system, never hitting the reflector surface. The phenomenon results in molding parameters (parts without stress), chemical resistance etc. a relatively large area of scattered light around the so-called hot spot area. Obviously, the angle can be minimized by increasing the height of The Future Is Bright the reflector, but it seldom is possible achieve a great extent of control. Finally, we want an answer to the ultimate question: why is it beneficial Reflectors often show a light distribution, which has a higher peak and to use plastic optics and why is a good optical solution the key to LED less smooth curve than what is the case for a lens of the same size. lighting revolution? The answers are many, but they all are simple and understandable: Combinations Price. No other material has such a high price/performance ratio as Combining reflectors and lenses in one optical system is a challenge, as optical plastic. Glass outperforms plastics in a few features, but overall, the optical functioning of them is very different and a combination the list for plastic is overwhelmingly long. easily results in a less efficient system than planned. Versatility. Plastics are the only optical materials, which can have Combining lenses in a system or doing the same with multiple reflectors numerous mechanical features integrated in the optical system. may be used with success. As an example of a flexible and cheap optical Different plastic materials can be used in different environments. system using lenses, LEDIL’s series of lenses use one and the same base Different plastics can be used to achieve different optical effects. lens that can be modified with add-on sub lenses, to change the illumination pattern. Performance. Only plastic optical systems have made LED flashlights possible in big volumes. The same can be said for most of the automotive System Thinking, Integration, Platforms LED applications – both interior and exterior. Without plastics and optics it would have been impossible to fulfill the requirements of these An important feature, when designing an illumination system, is to lighting systems. develop a modular system that can be easily adapted for different applications. When it comes to optical components, we have chosen a Starting in a big scale in a few years, we will all see a LED lighting few mechanical optics platforms, which we modify to their optical revolution take place in our homes, offices and public places. Take a geometry, but not to their mechanical dimensions. The end user can close look at all of the solutions. I can guarantee that you will see a feel confident that he always finds a standard component of a certain great number of plastic optics there, too, plastic optics – the enabler of size, with different optical characteristics and the same size and optical LED Light Revolution. choices for several, different LED types.

There are numerous ways that LEDs may be combined in series or series/ Driver parallel strings. Take, for example, a typical interior automotive lighting application requiring 200 lumens. Depending on LED selection, this requires a series string of three to six LEDs. Figure 1 shows a multi LED device package. To drive any LED combination that may be envisaged, New Drive Circuit Supports an efficient, high density, cost effective constant current converter Wide DC Input and Output with both a wide input (8V to 19V) and a wide output (6.9V to 30V) range is required. A pure buck or a pure boost switching regulator Range topology is not sufficiently flexible. What is proposed for the LED driver, therefore, is a current regulated, non inverting buck/boost converter. A > Bernie Weir, Director of Applications, ON Semiconductor high side current sensing scheme is used as it allows the LED string There are many reasons why lighting designers are looking to adopt LED cathode is connected directly to ground. Also, to maximize converter technology. Long operating life in the range of 50,000 hours, for performance, a sensing scheme having a low loss (e.g. 200mV) is example, helps to reduce ongoing maintenance costs, while instant required. The schematic of the LED driver is illustrated in Figure 2. turn on, even at temperatures as low as -40°C, is not possible with some other light sources. LEDs are small enough that they can be integrated into a wide variety of applications and, as a higher efficacy source of light than traditional incandescent, halogen and fluorescent, can help designers address commercial and environmental pressure to reduce power consumption. What’s more, there are ongoing advances in efficacy as illustrated in recent announcements from manufacturers such as Cree and Nichia of LEDs capable of delivering 130-150 lumens/ Watt in the lab. Finally, technology and manufacturing advances are helping the commercial argument for solid state lighting by continually driving down cost per lumen. In order to achieve specified brightness and color it is critical to drive LEDs with a constant current. For high brightness power LEDs currents Figure 2: Schematic of the new LED driver circuit. can range from 100mA to 1500mA, although 350mA is a common value. Table 1 shows manufacturers’ data for several product families, which has been used as the basis for creating minimum and maximum Theory of Operation voltage for a generic LED as shown. The output voltage swing for strings The simplified power stage is shown in Figure 3 for clarity. To minimize of three to six of these generic LEDs is also included for reference. From power dissipation in the power circuit, low ripple current is required. this data it is apparent that in normal operation variations of ±30% can So the converter is operated in a continuous current mode (CCM). be expected.

Figure 3: Simplified power stage of the driver.

For this analysis, all power components are assumed ideal. Switches Q1, Q2 turn on for time D*Ts (D duty cycle, Ts switching period) charging inductor L1 from input Vin. When Q1 and Q2 turn off, diodes D1, D2 deliver the inductor energy to the output Vout. For the inductor flux (V*μs) to remain in equilibrium each switching cycle, the V*μs product across the inductor during each switch interval must balance, as shown Figure 1: Example of a series string of LEDs in one package (OSRAM Ostar). in equation 1.

Rearranging equation 1, the voltage gain of buck boost is given by :

Varying the duty cycle will vary the output such that when D is below 0.5, the converter is in buck mode, when D is above 0.5, the converter is in boost mode and when D equals 0.5, the voltage gain Vout/Vin is unity. The ripple current in the inductor is given by expression:

Figure 5: Functional block diagram of the used controller, NCP3063. Assuming Vin = 12V and D*TS = 0.5*5μs, a value for L1 of 68μH in equation 3 will maintain +/-30% ripple current in a 700mA application This device consists of a 1.25V reference, comparator, oscillator, an active maintaining CCM operation. MOSFETs or BJTs can be selected as the current limit circuit, a driver and a high current output switch. In its primary switches Q1/Q2. NSS40500UW3T2G and NSS40501UW3T2G traditional operating mode, the NCP3063 is a hysteretic, dc - dc converter from ON Semiconductor’s e²PowerEdge family of BJTs were chosen for that uses a gated oscillator to regulate the output voltage. Voltage cost/performance criteria. They feature ultra low saturation voltage feedback from the output is sensed at pin 5, and gates the oscillator on/ and high current gain capability (Figure 4). off to regulate the output. The oscillator frequency and off-time of the output switch are programmed by the value selected for the timing capacitor, CT. CT is charged and discharged by a 1 to 6 ratio internal current source and sink, generating a ramp at pin 3. The ramp is controlled by two comparators whose levels are set 500mV apart. In normal operation, D is fixed at 6/7 or 0.86. The “gated oscillator” mode is used to protect the LED string if a LED fails ‘open”. A zener diode between Vout and pin 5 will clamp the output at a voltage VZ + 1.25V. The NCP3063 can also operate as a conventional PWM controller, by injecting current into the CT pin. The control current may be developed either from the input source, providing voltage feed-forward (via R5) or from the output current sensing circuit (via R6). In both cases the slope of the oscillator ramp changes causing D to vary. In Figure 2, the current sense resistor R9 is placed in series with Vout, to satisfy the high side sensing requirement. The bandgap reference U3, together with dual NPN transistors Q4a,b and R13, R14 create two equal current sinks. If U3 is a 1.25V bandgap and R13, R14 equal 1.24kΩ (1%) two 1mA current Figure 4: Saturation and current gain capability . sinks are formed. Resistors R10, R11 level shift the current sense signal IOUT*R9 to satisfy the input requirements of U2. To create a 210mV Turn on, turn off, saturation voltage and storage time of a BJT are reference for the current loop, the expression 1mA * (R10-R11) = 210mV controlled by the magnitudes of turn on IB1 and turn off IB2 base must be satisfied. Hence R10 is selected to be 210Ω larger in absolute currents. The drive currents are identified in Figure 3. The values of the value than R11. Current regulation is set by the equation Iout * R9 = base drive resistors R2, R3 and R4 in the schematic may be adjusted to 210mV. If R9 is 0.6Ω, the programmed current is set for 350mA. The optimize performance. The efficiency of the converter can be improved difference between the 210mV set point and the current sense is if the storage time of Q2 is less than Q1. The reasons for this will be amplified by U2 to create an error voltage. This error voltage and R6 discussed later. Q2’s storage time can be reduced if it is held out of drives a programmed current into the CT pin to regulate the LED saturation by the addition of D3a/b shown in Figure 3. Once Q2 is near current. saturation, additional base current flows through diode D3a and into the collector junction. This diversion of base current IB1 reduces the Because the converter is switching at 200kHz, MLCC capacitors in SMT stored charge in the base region and allows a faster turn off. Typical TS packages can provide cost effective filtering. Low value MLCC capacitors of 1.5μs is reduced to a few hundred nanoseconds. (10μF) have very small ESR (2mΩ) and ESL (100nH) values. When used

in single or parallel combinations they form a “perfect” capacitor. Ripple It is evident from Figures 6 and 7 that the inductor waveforms differ voltage is due only to charging and discharging the capacitor by the from a classic buck boost. The voltage across the inductor is clamped at inductor. Two 10μF, 1210 capacitors are employed across the input and (Vout–Vin) for the duration of the storage delay interval TD. During this output of the driver. The ripple voltage across the input capacitor = interval Q2 is off and Q1 is on for its storage time. During this period, D*Ts*ΔI (L1) / Cin. The ripple voltage developed across the output power is delivered to the output via Q1 and not by D1. Efficiency capacitor is given by (1-D)*Ts*ΔI (L1) / Cout. improvement is observed as the VCE(sat) of the PNP device (100mV) is less that the voltage drop across the Schottky diode D1 (300mV). If the Converter Waveforms time delay intervals were reversed and Q1 turned off first, power would cycle through the inductor L1, switch Q2 and diode D1. No power would The voltage waveforms at both the input (upper trace) and output be delivered to the load until Q2 turned off. The efficiency of the (lower trace) of the inductor L1 were measured while the difference converter is shown in figure 8 and varied between 75% and 80%. The waveform (middle trace) gives the voltage across the inductor. Figure 6 data was taken with Vin at 12V while the output was varied between 11 shows the converter operating in buck mode, while Figure 7 illustrates V and 26V at 700mA constant current load. boost operation. The V*μs balance expression given in equation 1 is modified as follows.

Conclusions ON Semiconductor’s latest monolithic NCP3063 controller and family of e2PowerEdge ultra low saturation bipolar transistors can be combined to create a non inverting buck boost topology optimized to drive strings of LEDs at a constant current. A high side, low drop, current sensing scheme has been implemented, targeted for automotive, marine, and other high efficiency applications were the input voltage may be loosely regulated. The output from the current sense is used to vary the slope of the oscillator ramp and achieve duty cycle modulation, independent of the gated oscillator function provided by the IC. The classic transfer function of the buck boost converter is modified by the storage time interval between the NPN and PNP bipolar switches.

Figure 6: Waveforms in buck mode operation.

Figure 7: Waveforms in boost mode operation.

test this concept. This buck topology is usually used with constant Simple and Accurate voltage circuits by referencing the output voltage through a zener back to the feedback pin. Output current is senses during the freewheeling Constant Current Sensing for action. When the MOSFET is off, the current circulates through L2, the Non-Isolated LED Drivers LEDs, D9 and R4. This period is longer than the on time of the MOSFET thus providing a reasonable accuracy. The current that is going through > John S. Lo Giudice, Application Engineer, STMicroelectronics R4 is turned into a voltage and is stored and filtered by C4. When the Many appliances and consumer electronics application are using LED MOSFET turns on again and D9 becomes back biased, the voltage that is lights that are driven by a non-isolated power supply. This circuit stored in C4 rides on top of the source, which is the reference for the IC, describes a novel approach to implement constant current control biasing Q2 which regulates the proper voltage to the feedback pin 3. mechanism inexpensively and accurately. The wide range of voltage for Vdd Pin, 9V to 38V, of this particular IC Convention approaches of implementing current limit would not work makes it ideal for this application, because the range of Vdd voltage is in non-isolated buck converters because of floating reference for the essentially the output voltage range. Hence this circuit can be used to control circuitry. In the circuit example shown below, a resistor and drive 3 to 10 LEDs. If the LED string is not connected, the output voltage capacitor is used in the freewheeling diode path to capture the falling will climb to the over voltage level and lead the circuit into a hick up edge of inductor current. Since freewheeling circuit carries current for mode. D6 blocks the voltage during the freewheeling time and can be a the majority of the duty cycle, this circuit provides reasonable accuracy. general purpose diode. In addition, this circuit solves the issue of floating ground by referencing Experimental results show that the circuit works well for a variable the control signal to same ground as the control circuitry. numbers of LEDs and can be even tighter for a fixed number. By changing A simple inexpensive Pulse Width modulation chip incorporated with a the number of LEDs, the load regulation can be checked. The line regulation MOSFET is ideal for building a constant current driver for LEDs. In the is also shown for a fixed number of LEDs from 90Vac to 264Vac. The circuit example shown below STMicroelectronics’ VIPer22A was used to results were measured on a board with the following results.

Figure 1. Schematic of circuit.

Since the current is triangular with a very large swing, C4 should be greater than 50 times the time constant. The higher the capacitance the better the regulation is. In this case 3Ω*330uF=9 ms which is much higher than the 17us period. Below is a scope picture of the current flowing through the inductor. The resistor turns the current into a voltage. The capacitor smooth out the peak voltage into an average voltage. That voltage is fed into the feed-back pin for regulation. The jump seen during the on time, which is represented by the rising of the current, is attributed to the common mode noise of the circuit because that point is flying from -1V to the peek of the input voltage. The regulation is derived from measured results on a power supply board by varying the load from 3 to 10, 1Watt LEDs.

Figure 3: Load regulation. Load regulation Line regulation LEDs Current Input 8 LEDs 3 362 90 350 4 358 115 329 5 352 132 326 6 344 180 329 7 336 230 336 8 328 264 340 9 321 10 314

Table 1: Load regulation, currents in mA (left), and Line regulation, voltages in Vac (right). Figure 4: Line regulation. This circuit delivers +/-6% load regulation from 4 to 10 LEDs and a line regulation of +/- 3.4% for 8 LEDs, when the line is varied from 90Vac to 264Vac.

Figure 5: Most recent driver board.

Figure 2: Inductor current and voltage for feedback. The actual board measures 55mm by 22mm making it compact for The above plotted currents and voltages give the following graphs. driving LEDs.

ADVERTISING & ARTICLES

Issue Editiorial Focus Space Close Material Close Pub.

Nov/Dec 2008 LED System Simulation and Testing Nov 7 Nov 14 Nov 28

Editorial Calendar 2008 LED professional Review (LpR)– Media Kit 2008 Send Email to: [email protected] www.led-professional.com Ref: LpR Media Kit or callLED us at:professional +43 5572 394489 Review | Sept/Oct 2008 | page 48 Copyright © 2008 Luger Research & LED-professional. All rights reserved.

The LED can be driven by PWM driving technique to maintain an average Reliability Test of LED Driven current through LED which is acceptable. The LED bulbs can resist high current stress [5] which means the LED can be driven directly from by PWM Technique battery by utilizing PWM technique. The present study focuses on > B. J. Huang et al., Department of Mechanical Engineering, reliability of LED luminaire using direct PWM driving technique. National Taiwan University Reliability Test Design The solar-powered LED lighting system has been commercialized for a The present study first setup a thermal chamber that is heated by two long time. The system usually consists of a DC to DC converter in order 100W tungsten lamps and uses two on/off to convert the battery voltage into a fixed voltage or current for the LED lighting luminary. This will cause energy loss and system reliability due controllers to control the chamber temperature to within 40±3oC. to the failure of DC/DC converter. In the present study, we develop a Figure 1 shows the design of the LED reliability test that is driven by the special technique to drive the LED luminary directly from battery direct PWM technique. A LED luminaire was specially designed for this utilizing PWM technique in order to remove the DC/DC converter. LED reliability test with four different circuits and each circuit connecting However, instantaneous current overdriven can occur easily due to the three LED lamps serially. variation of battery voltage with the state-of-charge of battery. In the present study, we setup a thermal chamber with temperature variation to within 40±3oC. A LED luminary was specially designed for the LED reliability test with four different circuits with each circuit connecting three LED lamps serially. A driver is designed to provide 4 kinds of power inputs to LED: (a) 350mA constant current, (b) 700mA,100Hz, duty cycle=50%, (c)700mA, 10KHz duty cycle=50% and (d) 1050mA, 100Hz, duty cycle=33%. The tests were performed simultaneously to compare light decay between normal drive condition (a) and other PWM driving conditions (b, c, d). The accumulated total test time so far is more then 7,032 hours and has shown no significant light decay in 4 different loops. This reveals that the PWM technique directly driven by battery is feasible and is able to reduce energy loss of DC to DC converter in the solar lighting system.

Introduction Figure 1: The design of the LED reliability test. The stand-alone-solar-powered lighting system must be properly designed in photovoltaic power generation capacity Wp, battery capacity C, and the power consumption, for obtaining a proper Lost of Special LED luminaire design Load Probability (LLP) to deal with rainy and cloudy days [1]. A good Figure 2 shows the aluminum PCB(Printed Circuit Board) and the special charge/discharge strategy is also needed for assuring the system LED luminaire design. The aluminum PCB (74mm diameter) has 4 circuits reliability [2]. This kind of solar-powered lighting system usually is (connecting 3 LEDs in series in each circuit). Each circuit was driven by installed in remote area where the grid power can not reach. It can save different driver. A thermocouple was soldered on the middle of the the power transmission line cost and has been shown that the stand- aluminum PCB for measuring the temperature of the aluminum PCB. A alone solar lighting system utilizing Light Emitting Diode (LED) as the heat conducting body was attached to the aluminum PCB to dissipate light source can save energy with reasonable payback time in remote the heat of aluminum PCB. The LEDs were selected from the same area [3]. The reason for utilizing LED is that the LED can have long product lot in a production line to reduce the variation of the thermal lifetime (about 50,000 hours when properly driven) and can be driven resistance in mass production. All LEDs were soldered on the aluminum by direct current (DC) from battery. But there is a problem for driving PCB. Therefore, each LED on the aluminum PCB has the same junction the LED directly from the battery. The battery voltage will change at temperature. Figure 3 shows the installation of a photo sensor (S2387) different state of charge and it may damage the LED. Therefore, the to measure the intensity of the light from LEDs. stand-alone solar LED lighting system usually consist a DC/DC converter to convert the battery voltage into the LED driving voltage [4]. The DC/ DC converter may solve the problem of the voltage variation but it also increases the system cost and reduce the system reliability due to additional component.

(a)350mA constant current: It is the most common way to drive the LED. This is the baseline of the LED lifetime. (b)700mA, Duty-Cycle=50%, PWM-Frequency=100Hz: In Case (b), the average current equal to 350mA but the current stress through the LED is two times greater then Case (a). (c)700mA, Duty-Cycle=50%, PWM-Frequency=10kHz: The average current equal to 350mA but the current through the LED Figure 2: The Aluminum PCB and special luminaire design. is two times greater then Case (a) condition and the PWM-Frequency is 100 times greater then (b). This test intends to see the effect of frequency and the pulse stress. (d)1050mA, Duty-Cycle=33%, PWM-Frequency=100Hz: The average current equals to 350mA but the current through the LED is three times greater then Case (a). Besides, a controller was designed using microprocessor to control the on/off for the different circuits of LEDs and measure the output of the photo sensor.

Figure 3: Installation design for photo sensor.

Thermal chamber The size of the thermal chamber is 60x90x60 cm made by 3cm thickness styrofoam. The chamber was heated by two 100W tungsten lamps. There are two on/off controllers to control the chamber temperature. Figure 4 shows the design and the test arrangement of the thermal chamber. Two block boards were placed in front of the two 100W Figure 4: Thermal chamber design and test arrangement. tungsten lamps for avoid heating the sensor of the on/off controller directly. Four thermocouples were placed at the same height and 10cm near the wall to measure the temperature distribution of the thermal chamber. Figure 5 shows the test result at four locations. The result shows that the thermal chamber can control the chamber temperature to within 40±3oC.

PWM driver and controller design A LED driver is designed to provide 4 kinds of power inputs to LED: (a) 350mA constant current, (b) 700mA, 100Hz, duty cycle=50%, (c)700mA, 10kHz duty cycle=50% and (d) 1050mA, 100Hz, duty Figure 5: The test of the thermal chamber. cycle=33%. The tests were performed simultaneously to compare different light decay between normal drive condition (a) and other PWM drive condition (b, c, d). Test Results The LED luminaire is put in the thermal chamber to accelerate the light Figure 6 shows the accumulated total test time so far is more then decay of the LED by increasing the ambient temperature. The power 7,032 hours and has shown no significant light decay in 4 different supply can provide a DC input to the driving controller then the driving circuits. The aluminum PCB temperature shown in Figure 6 remains in controller can generate 4 different driving current which are list below 55±3oC. to drive the LED.

Figure 6: LED reliability test result.

Conclusion The solar-powered LED lighting system usually consists of a DC to DC converter in order to convert the battery voltage into a fixed voltage or current for the LED luminaire input. This will cause energy loss and system reliability problem due to the failure of DC/DC converter. In the present study, we drive the LED luminaire directly from battery utilizing PWM technique to remove the DC/DC converter. However, instantaneous current overdriven can occur easily due to the variation of battery voltage with the state-of-charge of battery. A LED luminaire was specially designed for the LED reliability test with four different circuits with each circuit connecting three LED lamps serially. A driver is designed to provide 4 kinds of power inputs to LED: (a) 350mA constant current, (b) 700mA, 100Hz, duty cycle=50%, (c)700mA, 10kHz duty cycle=50% and (d) 1050mA, 100Hz, duty cycle=33%. The tests were performed simultaneously to compare light decay between normal drive condition (a) and other PWM driving conditions (b, c, d). The accumulated total test time so far is more then 7,032 hours and has shown no significant light decay in 4 different loops. This reveals that the PWM technique directly driven by battery is feasible and is able to reduce energy loss of DC to DC converter in the solar lighting system.

Acknowledgement: The present study was supported by Energy Bureau, Ministry of Economic Affairs, Taiwan.

References: [1] B. J. Huang, M. S. Wu, C. J. Wu, “Development and field test of a long-lasting solar LED lighting system“, Word Renewable Energy Congress IX August 19-25( 2006) p. 590. [2] IEEE Std.2362TM-2003, “IEEE Guide for Selection, Charging, Test, and Evaluation of Lead-Acid Batteries Used in Stand-Alone Photovoltaic (PV) System“ [3]B. J. Huang, Min-Sheng Wu, H. H. Huang and J. W. Chen, “Economic analysis of solar-powered LED roadway lighting“, Solar World Congress 2007, Sept. 17-22(2007) p. 466-470 [4] E. Koutroulis and K. Kalaitzakis, “Novel battery charging regulation system for photovoltaic applications“, IEE Proc.-Electr. Power Appl., Vol. 151, No. 2, March 2004 [5] M. Meneghini, S. Podda, A. Morelli, R. Pintus, L. Trevisanello, G. Meneghesso, M. Vanzi, E. Zanoni, “High brightness GaN LEDs degradation during dc and pulsed stress, “Microelectronics Reliability 46(2006), p.1720- 1724 [6] N. Narendran, Y. Gu, J. P. Freyssinier, H. Yu, L. Deng, “Solid-state lighting: failure analysis of white LEDs“, JOURNAL OF CRYSTAL GROWTH, 268(2004), p. 449-456

LED professional Review (LpR) Editorial Calendar 2009 Issue Main Feature Reservation by Files by Publication date

LED professional is a leading technology information source for Solid State Lighting, LED and OLED systems. LED professional reports in English language on latest lighting projects and products including optics, electronics, thermal management and controls. LED professional provides these information worldwide via the online portal (www.led-professional.com), a newsletter and the LED professional Review (LpR).

LED professional Online Portal

LED professional Newsletter

LED professional Review (LpR)

[email protected]

Publisher Luger Research Institute for Innovation & Technology dep LED professional Rhombergs Fabrik A 6850 Dornbirn, Austria | Europe phone +43 5572 394 489 faxwww.led-professional.co +43 5572 394 698 m LED professional Review | Sept/Oct 2008 | page 52 [email protected]

AZ_LED professional OK.indd 1 06.12.2007 9:31:01 Uhr